happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'ggml-org:master' into quant_types

This commit is contained in:
Bartowski 2025-12-15 12:38:55 -05:00 committed by GitHub
parent a312ac950d 4529c660c8
commit fe1f30b194
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
531 changed files with 129597 additions and 32490 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
.devops/cann.Dockerfile
View File

@ -4,7 +4,7 @@
# Define the CANN base image for easier version updates later
ARG CHIP_TYPE=910b
ARG CANN_BASE_IMAGE=quay.io/ascend/cann:8.3.rc1.alpha001-${CHIP_TYPE}-openeuler22.03-py3.11
ARG CANN_BASE_IMAGE=quay.io/ascend/cann:8.3.rc2-${CHIP_TYPE}-openeuler24.03-py3.11
# ==============================================================================
# BUILD STAGE
@ -111,7 +111,7 @@ ENTRYPOINT ["/app/tools.sh"]
# ==============================================================================
FROM base AS light
COPY --from=build /app/full/llama-cli /app
COPY --from=build /app/full/llama-cli /app/full/llama-completion /app
ENTRYPOINT [ "/app/llama-cli" ]

2
.devops/cpu.Dockerfile
View File

@ -68,7 +68,7 @@ ENTRYPOINT ["/app/tools.sh"]
### Light, CLI only
FROM base AS light
COPY --from=build /app/full/llama-cli /app
COPY --from=build /app/full/llama-cli /app/full/llama-completion /app
WORKDIR /app

2
.devops/cuda.Dockerfile
View File

@ -74,7 +74,7 @@ ENTRYPOINT ["/app/tools.sh"]
### Light, CLI only
FROM base AS light
COPY --from=build /app/full/llama-cli /app
COPY --from=build /app/full/llama-cli /app/full/llama-completion /app
WORKDIR /app

2
.devops/intel.Dockerfile
View File

@ -73,7 +73,7 @@ ENTRYPOINT ["/app/tools.sh"]
FROM base AS light
COPY --from=build /app/lib/ /app
COPY --from=build /app/full/llama-cli /app
COPY --from=build /app/full/llama-cli /app/full/llama-completion /app
WORKDIR /app

2
.devops/musa.Dockerfile
View File

@ -81,7 +81,7 @@ ENTRYPOINT ["/app/tools.sh"]
### Light, CLI only
FROM base AS light
COPY --from=build /app/full/llama-cli /app
COPY --from=build /app/full/llama-cli /app/full/llama-completion /app
WORKDIR /app

2
.devops/rocm.Dockerfile
View File

@ -94,7 +94,7 @@ ENTRYPOINT ["/app/tools.sh"]
### Light, CLI only
FROM base AS light
COPY --from=build /app/full/llama-cli /app
COPY --from=build /app/full/llama-cli /app/full/llama-completion /app
WORKDIR /app

2
.devops/s390x.Dockerfile
View File

@ -105,7 +105,7 @@ WORKDIR /llama.cpp/bin
# Copy llama.cpp binaries and libraries
COPY --from=collector /llama.cpp/bin/*.so /llama.cpp/bin
COPY --from=collector /llama.cpp/bin/llama-cli /llama.cpp/bin
COPY --from=collector /llama.cpp/bin/llama-cli /llama.cpp/bin/llama-completion /llama.cpp/bin
ENTRYPOINT [ "/llama.cpp/bin/llama-cli" ]

8
.devops/tools.sh
View File

@ -13,6 +13,8 @@ elif [[ "$arg1" == '--quantize' || "$arg1" == '-q' ]]; then
exec ./llama-quantize "$@"
elif [[ "$arg1" == '--run' || "$arg1" == '-r' ]]; then
exec ./llama-cli "$@"
elif [[ "$arg1" == '--run-legacy' || "$arg1" == '-l' ]]; then
exec ./llama-completion "$@"
elif [[ "$arg1" == '--bench' || "$arg1" == '-b' ]]; then
exec ./llama-bench "$@"
elif [[ "$arg1" == '--perplexity' || "$arg1" == '-p' ]]; then
@ -32,8 +34,10 @@ elif [[ "$arg1" == '--server' || "$arg1" == '-s' ]]; then
else
echo "Unknown command: $arg1"
echo "Available commands: "
echo " --run (-r): Run a model previously converted into ggml"
echo " ex: -m /models/7B/ggml-model-q4_0.bin -p \"Building a website can be done in 10 simple steps:\" -n 512"
echo " --run (-r): Run a model (chat) previously converted into ggml"
echo " ex: -m /models/7B/ggml-model-q4_0.bin"
echo " --run-legacy (-l): Run a model (legacy completion) previously converted into ggml"
echo " ex: -m /models/7B/ggml-model-q4_0.bin -no-cnv -p \"Building a website can be done in 10 simple steps:\" -n 512"
echo " --bench (-b): Benchmark the performance of the inference for various parameters."
echo " ex: -m model.gguf"
echo " --perplexity (-p): Measure the perplexity of a model over a given text."

2
.devops/vulkan.Dockerfile
View File

@ -68,7 +68,7 @@ ENTRYPOINT ["/app/tools.sh"]
### Light, CLI only
FROM base AS light
COPY --from=build /app/full/llama-cli /app
COPY --from=build /app/full/llama-cli /app/full/llama-completion /app
WORKDIR /app

9
.github/ISSUE_TEMPLATE/011-bug-results.yml vendored
View File

@ -11,7 +11,7 @@ body:
(i.e. the generated text) are incorrect or llama.cpp crashes during model evaluation.
If you encountered the issue while using an external UI (e.g. ollama),
please reproduce your issue using one of the examples/binaries in this repository.
The `llama-cli` binary can be used for simple and reproducible model inference.
The `llama-completion` binary can be used for simple and reproducible model inference.
- type: textarea
id: version
attributes:
@ -74,9 +74,12 @@ body:
Please give us a summary of the problem and tell us how to reproduce it.
If you can narrow down the bug to specific hardware, compile flags, or command line arguments,
that information would be very much appreciated by us.
If possible, please try to reproduce the issue using `llama-completion` with `-fit off`.
If you can only reproduce the issue with `-fit on`, please provide logs both with and without `--verbose`.
placeholder: >
e.g. when I run llama-cli with -ngl 99 I get garbled outputs.
When I use -ngl 0 it works correctly.
e.g. when I run llama-completion with `-fa on` I get garbled outputs for very long prompts.
With short prompts or `-fa off` it works correctly.
Here are the exact commands that I used: ...
validations:
required: true

31
.github/actions/windows-setup-cuda/action.yml vendored
View File

@ -65,3 +65,34 @@ runs:
echo "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4\libnvvp" | Out-File -FilePath $env:GITHUB_PATH -Encoding utf8 -Append
echo "CUDA_PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4" | Out-File -FilePath $env:GITHUB_ENV -Append -Encoding utf8
echo "CUDA_PATH_V12_4=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4" | Out-File -FilePath $env:GITHUB_ENV -Append -Encoding utf8
- name: Install Cuda Toolkit 13.1
if: ${{ inputs.cuda_version == '13.1' }}
shell: pwsh
run: |
mkdir -p "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1"
choco install unzip -y
curl -O "https://developer.download.nvidia.com/compute/cuda/redist/cuda_crt/windows-x86_64/cuda_crt-windows-x86_64-13.1.80-archive.zip"
curl -O "https://developer.download.nvidia.com/compute/cuda/redist/cuda_cudart/windows-x86_64/cuda_cudart-windows-x86_64-13.1.80-archive.zip"
curl -O "https://developer.download.nvidia.com/compute/cuda/redist/cuda_nvcc/windows-x86_64/cuda_nvcc-windows-x86_64-13.1.80-archive.zip"
curl -O "https://developer.download.nvidia.com/compute/cuda/redist/cuda_nvrtc/windows-x86_64/cuda_nvrtc-windows-x86_64-13.1.80-archive.zip"
curl -O "https://developer.download.nvidia.com/compute/cuda/redist/libcublas/windows-x86_64/libcublas-windows-x86_64-13.2.0.9-archive.zip"
curl -O "https://developer.download.nvidia.com/compute/cuda/redist/libnvvm/windows-x86_64/libnvvm-windows-x86_64-13.1.80-archive.zip"
curl -O "https://developer.download.nvidia.com/compute/cuda/redist/cuda_nvtx/windows-x86_64/cuda_nvtx-windows-x86_64-13.1.68-archive.zip"
curl -O "https://developer.download.nvidia.com/compute/cuda/redist/cuda_profiler_api/windows-x86_64/cuda_profiler_api-windows-x86_64-13.1.80-archive.zip"
curl -O "https://developer.download.nvidia.com/compute/cuda/redist/visual_studio_integration/windows-x86_64/visual_studio_integration-windows-x86_64-13.1.68-archive.zip"
curl -O "https://developer.download.nvidia.com/compute/cuda/redist/cuda_cccl/windows-x86_64/cuda_cccl-windows-x86_64-13.1.78-archive.zip"
unzip '*.zip' -d "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1"
xcopy "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1\cuda_crt-windows-x86_64-13.1.80-archive\*" "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1" /E /I /H /Y
xcopy "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1\cuda_cudart-windows-x86_64-13.1.80-archive\*" "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1" /E /I /H /Y
xcopy "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1\cuda_nvcc-windows-x86_64-13.1.80-archive\*" "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1" /E /I /H /Y
xcopy "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1\cuda_nvrtc-windows-x86_64-13.1.80-archive\*" "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1" /E /I /H /Y
xcopy "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1\libcublas-windows-x86_64-13.2.0.9-archive\*" "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1" /E /I /H /Y
xcopy "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1\libnvvm-windows-x86_64-13.1.80-archive\*" "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1" /E /I /H /Y
xcopy "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1\cuda_nvtx-windows-x86_64-13.1.68-archive\*" "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1" /E /I /H /Y
xcopy "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1\cuda_profiler_api-windows-x86_64-13.1.80-archive\*" "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1" /E /I /H /Y
xcopy "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1\visual_studio_integration-windows-x86_64-13.1.68-archive\*" "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1" /E /I /H /Y
xcopy "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1\cuda_cccl-windows-x86_64-13.1.78-archive\*" "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1" /E /I /H /Y
echo "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1\bin" | Out-File -FilePath $env:GITHUB_PATH -Encoding utf8 -Append
echo "CUDA_PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1" | Out-File -FilePath $env:GITHUB_ENV -Append -Encoding utf8
echo "CUDA_PATH_V13_1=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.1" | Out-File -FilePath $env:GITHUB_ENV -Append -Encoding utf8

1
.github/workflows/build-linux-cross.yml vendored
View File

@ -291,6 +291,7 @@ jobs:
-DGGML_RVV=ON \
-DGGML_RV_ZFH=ON \
-DGGML_RV_ZICBOP=ON \
-DGGML_RV_ZIHINTPAUSE=ON \
-DRISCV64_SPACEMIT_IME_SPEC=RISCV64_SPACEMIT_IME1 \
-DCMAKE_TOOLCHAIN_FILE=${PWD}/cmake/riscv64-spacemit-linux-gnu-gcc.cmake

120
.github/workflows/build-riscv-native.yml vendored
View File

@ -1,120 +0,0 @@
name: Build on RISCV Linux Machine by Cloud-V
on:
pull_request:
workflow_dispatch:
workflow_call:
jobs:
debian-13-riscv64-native: # Bianbu 2.2
runs-on: [self-hosted, RISCV64]
steps:
- name: Install prerequisites
run: |
sudo apt-get update || true
sudo apt-get install -y libatomic1
- uses: actions/checkout@v4
- name: Setup Riscv
run: |
sudo apt-get update || true
sudo apt-get install -y --no-install-recommends \
build-essential \
gcc-14-riscv64-linux-gnu \
g++-14-riscv64-linux-gnu \
ccache \
cmake
- name: Setup ccache
run: |
mkdir -p $HOME/.ccache
ccache -M 5G -d $HOME/.ccache
export CCACHE_LOGFILE=/home/runneruser/ccache_debug/ccache.log
export CCACHE_DEBUGDIR="/home/runneruser/ccache_debug"
echo "$GITHUB_WORKSPACE"
echo "CCACHE_LOGFILE=$CCACHE_LOGFILE" >> $GITHUB_ENV
echo "CCACHE_DEBUGDIR=$CCACHE_DEBUGDIR" >> $GITHUB_ENV
echo "CCACHE_BASEDIR=$GITHUB_WORKSPACE" >> $GITHUB_ENV
echo "CCACHE_DIR=$HOME/.ccache" >> $GITHUB_ENV
- name: Build
run: |
cmake -B build \
-DLLAMA_CURL=OFF \
-DCMAKE_BUILD_TYPE=Release \
-DGGML_OPENMP=OFF \
-DLLAMA_BUILD_EXAMPLES=ON \
-DLLAMA_BUILD_TOOLS=ON \
-DLLAMA_BUILD_TESTS=OFF \
-DCMAKE_SYSTEM_NAME=Linux \
-DCMAKE_SYSTEM_PROCESSOR=riscv64 \
-DCMAKE_C_COMPILER=riscv64-linux-gnu-gcc-14 \
-DCMAKE_CXX_COMPILER=riscv64-linux-gnu-g++-14 \
-DCMAKE_C_COMPILER_LAUNCHER=ccache \
-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
-DCMAKE_POSITION_INDEPENDENT_CODE=ON \
-DCMAKE_FIND_ROOT_PATH=/usr/lib/riscv64-linux-gnu \
-DCMAKE_FIND_ROOT_PATH_MODE_PROGRAM=NEVER \
-DCMAKE_FIND_ROOT_PATH_MODE_LIBRARY=ONLY \
-DCMAKE_FIND_ROOT_PATH_MODE_INCLUDE=BOTH
cmake --build build --config Release -j $(nproc)
# debian-13-riscv64-spacemit-ime-native: # Bianbu 2.2
# runs-on: [self-hosted, RISCV64]
# steps:
# - name: Install prerequisites
# run: |
# sudo apt-get update || true
# sudo apt-get install -y libatomic1
# - uses: actions/checkout@v4
# - name: Setup Riscv
# run: |
# sudo apt-get update || true
# sudo apt-get install -y --no-install-recommends \
# build-essential \
# gcc-14-riscv64-linux-gnu \
# g++-14-riscv64-linux-gnu \
# ccache \
# cmake
# sudo apt-get upgrade binutils -y
# - name: Setup ccache
# run: |
# mkdir -p $HOME/.ccache
# ccache -M 5G -d $HOME/.ccache
# export CCACHE_LOGFILE=/home/runneruser/ccache_debug/ccache.log
# export CCACHE_DEBUGDIR="/home/runneruser/ccache_debug"
# echo "$GITHUB_WORKSPACE"
# echo "CCACHE_LOGFILE=$CCACHE_LOGFILE" >> $GITHUB_ENV
# echo "CCACHE_DEBUGDIR=$CCACHE_DEBUGDIR" >> $GITHUB_ENV
# echo "CCACHE_BASEDIR=$GITHUB_WORKSPACE" >> $GITHUB_ENV
# echo "CCACHE_DIR=$HOME/.ccache" >> $GITHUB_ENV
# - name: Build
# run: |
# cmake -B build \
# -DLLAMA_CURL=OFF \
# -DCMAKE_BUILD_TYPE=Release \
# -DGGML_OPENMP=OFF \
# -DLLAMA_BUILD_EXAMPLES=ON \
# -DLLAMA_BUILD_TOOLS=ON \
# -DLLAMA_BUILD_TESTS=OFF \
# -DCMAKE_SYSTEM_NAME=Linux \
# -DCMAKE_SYSTEM_PROCESSOR=riscv64 \
# -DCMAKE_C_COMPILER=riscv64-linux-gnu-gcc-14 \
# -DCMAKE_CXX_COMPILER=riscv64-linux-gnu-g++-14 \
# -DCMAKE_C_COMPILER_LAUNCHER=ccache \
# -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
# -DCMAKE_POSITION_INDEPENDENT_CODE=ON \
# -DCMAKE_FIND_ROOT_PATH=/usr/lib/riscv64-linux-gnu \
# -DCMAKE_FIND_ROOT_PATH_MODE_PROGRAM=NEVER \
# -DCMAKE_FIND_ROOT_PATH_MODE_LIBRARY=ONLY \
# -DCMAKE_FIND_ROOT_PATH_MODE_INCLUDE=BOTH \
# -DGGML_RVV=ON \
# -DGGML_RV_ZFH=ON \
# -DGGML_RV_ZICBOP=ON \
# -DGGML_CPU_RISCV64_SPACEMIT=ON \
# -DRISCV64_SPACEMIT_IME_SPEC=RISCV64_SPACEMIT_IME1
# cmake --build build --config Release -j $(nproc)

474
.github/workflows/build.yml vendored
View File

@ -20,7 +20,8 @@ on:
'**/*.swift',
'**/*.m',
'**/*.metal',
'**/*.comp'
'**/*.comp',
'**/*.glsl'
]
pull_request:
@ -40,7 +41,8 @@ on:
'**/*.swift',
'**/*.m',
'**/*.metal',
'**/*.comp'
'**/*.comp',
'**/*.glsl'
]
concurrency:
@ -243,7 +245,7 @@ jobs:
echo "Fetch llama2c model"
wget https://huggingface.co/karpathy/tinyllamas/resolve/main/stories260K/stories260K.bin
./bin/llama-convert-llama2c-to-ggml --copy-vocab-from-model ./tok512.bin --llama2c-model stories260K.bin --llama2c-output-model stories260K.gguf
./bin/llama-cli -m stories260K.gguf -p "One day, Lily met a Shoggoth" -n 500 -c 256
./bin/llama-completion -m stories260K.gguf -p "One day, Lily met a Shoggoth" -n 500 -c 256
- name: Test llama2c (s390x)
id: llama2c_test_s390x
@ -252,7 +254,7 @@ jobs:
cd build
echo "Fetch llama2c big-endian model"
wget https://huggingface.co/ggml-org/models/resolve/main/tinyllamas/stories260K-be.gguf
./bin/llama-cli -m stories260K-be.gguf -p "One day, Lily met a Shoggoth" -n 500 -c 256
./bin/llama-completion -m stories260K-be.gguf -p "One day, Lily met a Shoggoth" -n 500 -c 256
ubuntu-latest-cmake-sanitizer:
runs-on: ubuntu-latest
@ -547,6 +549,46 @@ jobs:
# This is using llvmpipe and runs slower than other backends
ctest -L main --verbose --timeout 3600
ubuntu-24-wasm-webgpu:
runs-on: ubuntu-24.04
steps:
- name: Clone
id: checkout
uses: actions/checkout@v4
- name: ccache
uses: ggml-org/ccache-action@v1.2.16
with:
key: ubuntu-latest-wasm-webgpu
evict-old-files: 1d
- name: Install Emscripten
run: |
git clone https://github.com/emscripten-core/emsdk.git
cd emsdk
./emsdk install latest
./emsdk activate latest
- name: Fetch emdawnwebgpu
run: |
DAWN_TAG="v20251027.212519"
EMDAWN_PKG="emdawnwebgpu_pkg-${DAWN_TAG}.zip"
echo "Downloading ${EMDAWN_PKG}"
curl -L -o emdawn.zip \
"https://github.com/google/dawn/releases/download/${DAWN_TAG}/${EMDAWN_PKG}"
unzip emdawn.zip
- name: Build WASM WebGPU
run: |
source emsdk/emsdk_env.sh
emcmake cmake -B build-wasm \
-DGGML_WEBGPU=ON \
-DLLAMA_CURL=OFF \
-DEMDAWNWEBGPU_DIR=emdawnwebgpu_pkg
cmake --build build-wasm --target test-backend-ops -j $(nproc)
ubuntu-22-cmake-hip:
runs-on: ubuntu-22.04
container: rocm/dev-ubuntu-22.04:6.1.2
@ -1360,25 +1402,54 @@ jobs:
chip_type: ['910b', '310p']
build: ['Release']
runs-on: ${{ matrix.arch == 'aarch64' && 'ubuntu-24.04-arm' || 'ubuntu-24.04' }}
container: ascendai/cann:${{ matrix.chip_type == '910b' && '8.3.rc1.alpha001-910b-openeuler22.03-py3.11' || '8.2.rc1-310p-openeuler22.03-py3.11' }}
steps:
- name: Checkout
uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Dependencies
- name: Free up disk space
uses: ggml-org/free-disk-space@v1.3.1
with:
tool-cache: true
- name: Set container image
id: cann-image
run: |
yum update -y
yum install -y git gcc gcc-c++ make cmake libcurl-devel
image="ascendai/cann:${{ matrix.chip_type == '910b' && '8.3.rc2-910b-openeuler24.03-py3.11' || '8.3.rc2-310p-openeuler24.03-py3.11' }}"
echo "image=${image}" >> "${GITHUB_OUTPUT}"
- name: Pull container image
run: docker pull "${{ steps.cann-image.outputs.image }}"
- name: Build
env:
BUILD_TYPE: ${{ matrix.build }}
SOC_TYPE: ascend${{ matrix.chip_type }}
run: |
export LD_LIBRARY_PATH=${ASCEND_TOOLKIT_HOME}/lib64:${ASCEND_TOOLKIT_HOME}/$(uname -m)-linux/devlib/:${LD_LIBRARY_PATH}
HOST_UID=$(id -u)
HOST_GID=$(id -g)
cmake -S . -B build \
-DCMAKE_BUILD_TYPE=${{ matrix.build }} \
-DGGML_CANN=on \
-DSOC_TYPE=ascend${{ matrix.chip_type }}
cmake --build build -j $(nproc)
docker run --rm \
-v "${PWD}:/workspace" \
-w /workspace \
-e SOC_TYPE=${SOC_TYPE} \
-e BUILD_TYPE=${BUILD_TYPE} \
"${{ steps.cann-image.outputs.image }}" \
bash -lc '
set -e
yum install -y --setopt=install_weak_deps=False --setopt=tsflags=nodocs git gcc gcc-c++ make cmake libcurl-devel
yum clean all && rm -rf /var/cache/yum
git config --global --add safe.directory "/workspace"
export LD_LIBRARY_PATH=${ASCEND_TOOLKIT_HOME}/lib64:${ASCEND_TOOLKIT_HOME}/$(uname -m)-linux/devlib/:${LD_LIBRARY_PATH}
cmake -S . -B build \
-DCMAKE_BUILD_TYPE=${BUILD_TYPE} \
-DGGML_CANN=on \
-DSOC_TYPE=${SOC_TYPE}
cmake --build build -j $(nproc)
chown -R '"${HOST_UID}"':'"${HOST_GID}"' /workspace/build
'
# TODO: simplify the following workflows using a matrix
# TODO: run lighter CI on PRs and the full CI only on master (if needed)
@ -1562,33 +1633,33 @@ jobs:
run: |
bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
ggml-ci-x64-amd-vulkan:
runs-on: [self-hosted, Linux, X64, AMD]
# ggml-ci-x64-amd-vulkan:
# runs-on: [self-hosted, Linux, X64, AMD]
steps:
- name: Clone
id: checkout
uses: actions/checkout@v4
# steps:
# - name: Clone
# id: checkout
# uses: actions/checkout@v4
- name: Test
id: ggml-ci
run: |
vulkaninfo --summary
GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
# - name: Test
# id: ggml-ci
# run: |
# vulkaninfo --summary
# GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
ggml-ci-x64-amd-rocm:
runs-on: [self-hosted, Linux, X64, AMD]
# ggml-ci-x64-amd-rocm:
# runs-on: [self-hosted, Linux, X64, AMD]
steps:
- name: Clone
id: checkout
uses: actions/checkout@v4
# steps:
# - name: Clone
# id: checkout
# uses: actions/checkout@v4
- name: Test
id: ggml-ci
run: |
amd-smi static
GG_BUILD_ROCM=1 GG_BUILD_AMDGPU_TARGETS="gfx1101" bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
# - name: Test
# id: ggml-ci
# run: |
# amd-smi static
# GG_BUILD_ROCM=1 GG_BUILD_AMDGPU_TARGETS="gfx1101" bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
ggml-ci-mac-metal:
runs-on: [self-hosted, macOS, ARM64]
@ -1642,6 +1713,337 @@ jobs:
run: |
GG_BUILD_KLEIDIAI=1 GG_BUILD_EXTRA_TESTS_0=1 bash ./ci/run.sh ./tmp/results ./tmp/mnt
ubuntu-cpu-cmake-riscv64-native:
runs-on: RISCV64
steps:
- name: Install dependencies
run: |
sudo apt-get update
# Install necessary packages
sudo apt-get install -y libatomic1 libtsan2 gcc-14 g++-14 rustup cmake build-essential libssl-dev wget ccache
# Set gcc-14 and g++-14 as the default compilers
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-14 100
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-14 100
sudo ln -sf /usr/bin/gcc-14 /usr/bin/gcc
sudo ln -sf /usr/bin/g++-14 /usr/bin/g++
# Install Rust stable version
rustup install stable
rustup default stable
- name: Clone
id: checkout
uses: actions/checkout@v4
- name: Check environment
run: |
uname -a
gcc --version
g++ --version
ldd --version
cmake --version
rustc --version
- name: Setup ccache
run: |
# Set unique cache directory for this job
export CCACHE_DIR="$HOME/.ccache/cpu-cmake-rv64-native"
mkdir -p "$CCACHE_DIR"
# Configure ccache for optimal performance
ccache --set-config=max_size=5G
ccache --set-config=compression=true
ccache --set-config=compression_level=6
ccache --set-config=cache_dir="$CCACHE_DIR"
# Enable more aggressive caching
ccache --set-config=sloppiness=file_macro,time_macros,include_file_mtime,include_file_ctime
ccache --set-config=hash_dir=false
# Export for subsequent steps
echo "CCACHE_DIR=$CCACHE_DIR" >> $GITHUB_ENV
echo "PATH=/usr/lib/ccache:$PATH" >> $GITHUB_ENV
- name: Build
id: cmake_build
run: |
cmake -B build \
-DLLAMA_CURL=OFF \
-DLLAMA_OPENSSL=ON \
-DCMAKE_BUILD_TYPE=Release \
-DGGML_OPENMP=OFF \
-DLLAMA_BUILD_EXAMPLES=ON \
-DLLAMA_BUILD_TOOLS=ON \
-DLLAMA_BUILD_TESTS=ON \
-DCMAKE_C_COMPILER_LAUNCHER=ccache \
-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
-DGGML_RPC=ON \
-DCMAKE_C_COMPILER=riscv64-linux-gnu-gcc-14 \
-DCMAKE_CXX_COMPILER=riscv64-linux-gnu-g++-14
cmake --build build --config Release -j $(nproc)
- name: Test
id: cmake_test
run: |
cd build
ctest -L 'main|curl' --verbose --timeout 900
- name: Test llama2c conversion
id: llama2c_test
run: |
cd build
echo "Fetch tokenizer"
wget https://huggingface.co/karpathy/tinyllamas/resolve/main/stories260K/tok512.bin
echo "Fetch llama2c model"
wget https://huggingface.co/karpathy/tinyllamas/resolve/main/stories260K/stories260K.bin
./bin/llama-convert-llama2c-to-ggml --copy-vocab-from-model ./tok512.bin --llama2c-model stories260K.bin --llama2c-output-model stories260K.gguf
./bin/llama-completion -m stories260K.gguf -p "One day, Lily met a Shoggoth" -n 500 -c 256
ubuntu-cmake-sanitizer-riscv64-native:
runs-on: RISCV64
continue-on-error: true
strategy:
matrix:
sanitizer: [ADDRESS, THREAD, UNDEFINED]
build_type: [Debug]
steps:
- name: Install dependencies
run: |
sudo apt-get update
# Install necessary packages
sudo apt-get install -y libatomic1 libtsan2 gcc-14 g++-14 rustup cmake build-essential wget ccache
# Set gcc-14 and g++-14 as the default compilers
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-14 100
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-14 100
sudo ln -sf /usr/bin/gcc-14 /usr/bin/gcc
sudo ln -sf /usr/bin/g++-14 /usr/bin/g++
# Install Rust stable version
rustup install stable
rustup default stable
- name: GCC version check
run: |
gcc --version
g++ --version
- name: Clone
id: checkout
uses: actions/checkout@v4
- name: Setup ccache
run: |
# Unique cache directory per matrix combination
export CCACHE_DIR="$HOME/.ccache/sanitizer-${{ matrix.sanitizer }}-${{ matrix.build_type }}"
mkdir -p "$CCACHE_DIR"
# Configure ccache
ccache --set-config=max_size=5G
ccache --set-config=compression=true
ccache --set-config=compression_level=6
ccache --set-config=cache_dir="$CCACHE_DIR"
ccache --set-config=sloppiness=file_macro,time_macros,include_file_mtime,include_file_ctime
ccache --set-config=hash_dir=false
# Export for subsequent steps
echo "CCACHE_DIR=$CCACHE_DIR" >> $GITHUB_ENV
echo "PATH=/usr/lib/ccache:$PATH" >> $GITHUB_ENV
- name: Build
id: cmake_build
if: ${{ matrix.sanitizer != 'THREAD' }}
run: |
cmake -B build \
-DLLAMA_CURL=OFF \
-DCMAKE_BUILD_TYPE=${{ matrix.build_type }} \
-DGGML_OPENMP=ON \
-DLLAMA_BUILD_EXAMPLES=ON \
-DLLAMA_BUILD_TOOLS=ON \
-DLLAMA_BUILD_TESTS=OFF \
-DCMAKE_C_COMPILER_LAUNCHER=ccache \
-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
-DLLAMA_SANITIZE_${{ matrix.sanitizer }}=ON \
-DCMAKE_C_COMPILER=riscv64-linux-gnu-gcc-14 \
-DCMAKE_CXX_COMPILER=riscv64-linux-gnu-g++-14
cmake --build build --config ${{ matrix.build_type }} -j $(nproc)
- name: Build (no OpenMP)
id: cmake_build_no_openmp
if: ${{ matrix.sanitizer == 'THREAD' }}
run: |
cmake -B build \
-DLLAMA_CURL=OFF \
-DCMAKE_BUILD_TYPE=${{ matrix.build_type }} \
-DGGML_OPENMP=OFF \
-DLLAMA_BUILD_EXAMPLES=ON \
-DLLAMA_BUILD_TOOLS=ON \
-DLLAMA_BUILD_TESTS=OFF \
-DCMAKE_C_COMPILER_LAUNCHER=ccache \
-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
-DLLAMA_SANITIZE_${{ matrix.sanitizer }}=ON \
-DCMAKE_C_COMPILER=riscv64-linux-gnu-gcc-14 \
-DCMAKE_CXX_COMPILER=riscv64-linux-gnu-g++-14
cmake --build build --config ${{ matrix.build_type }} -j $(nproc)
- name: Test
id: cmake_test
run: |
cd build
ctest -L main --verbose --timeout 900
ubuntu-llguidance-riscv64-native:
runs-on: RISCV64
steps:
- name: Install dependencies
run: |
sudo apt-get update
# Install necessary packages
sudo apt-get install -y libatomic1 libtsan2 gcc-14 g++-14 rustup cmake build-essential wget ccache
# Set gcc-14 and g++-14 as the default compilers
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-14 100
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-14 100
sudo ln -sf /usr/bin/gcc-14 /usr/bin/gcc
sudo ln -sf /usr/bin/g++-14 /usr/bin/g++
# Install Rust stable version
rustup install stable
rustup default stable
- name: GCC version check
run: |
gcc --version
g++ --version
- name: Clone
id: checkout
uses: actions/checkout@v4
- name: Setup ccache
run: |
export CCACHE_DIR="$HOME/.ccache/llguidance-riscv64"
mkdir -p "$CCACHE_DIR"
ccache --set-config=max_size=5G
ccache --set-config=compression=true
ccache --set-config=compression_level=6
ccache --set-config=cache_dir="$CCACHE_DIR"
ccache --set-config=sloppiness=file_macro,time_macros,include_file_mtime,include_file_ctime
ccache --set-config=hash_dir=false
echo "CCACHE_DIR=$CCACHE_DIR" >> $GITHUB_ENV
echo "PATH=/usr/lib/ccache:$PATH" >> $GITHUB_ENV
- name: Build
id: cmake_build
run: |
cmake -B build \
-DLLAMA_CURL=OFF \
-DCMAKE_BUILD_TYPE=Release \
-DGGML_OPENMP=OFF \
-DLLAMA_BUILD_EXAMPLES=ON \
-DLLAMA_BUILD_TOOLS=ON \
-DLLAMA_BUILD_TESTS=OFF \
-DCMAKE_C_COMPILER_LAUNCHER=ccache \
-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
-DLLAMA_LLGUIDANCE=ON \
-DCMAKE_C_COMPILER=riscv64-linux-gnu-gcc-14 \
-DCMAKE_CXX_COMPILER=riscv64-linux-gnu-g++-14
cmake --build build --config Release -j $(nproc)
- name: Test
id: cmake_test
run: |
cd build
ctest -L main --verbose --timeout 900
ubuntu-cmake-rpc-riscv64-native:
runs-on: RISCV64
continue-on-error: true
steps:
- name: Install dependencies
run: |
sudo apt-get update
# Install necessary packages
sudo apt-get install -y libatomic1 libtsan2 gcc-14 g++-14 rustup cmake build-essential libssl-dev wget ccache
# Set gcc-14 and g++-14 as the default compilers
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-14 100
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-14 100
sudo ln -sf /usr/bin/gcc-14 /usr/bin/gcc
sudo ln -sf /usr/bin/g++-14 /usr/bin/g++
# Install Rust stable version
rustup install stable
rustup default stable
- name: GCC version check
run: |
gcc --version
g++ --version
- name: Clone
id: checkout
uses: actions/checkout@v4
- name: Setup ccache
run: |
export CCACHE_DIR="$HOME/.ccache/rpc-riscv64"
mkdir -p "$CCACHE_DIR"
ccache --set-config=max_size=5G
ccache --set-config=compression=true
ccache --set-config=compression_level=6
ccache --set-config=cache_dir="$CCACHE_DIR"
ccache --set-config=sloppiness=file_macro,time_macros,include_file_mtime,include_file_ctime
ccache --set-config=hash_dir=false
echo "CCACHE_DIR=$CCACHE_DIR" >> $GITHUB_ENV
echo "PATH=/usr/lib/ccache:$PATH" >> $GITHUB_ENV
- name: Build
id: cmake_build
run: |
cmake -B build \
-DLLAMA_CURL=OFF \
-DLLAMA_OPENSSL=ON \
-DCMAKE_BUILD_TYPE=Release \
-DGGML_OPENMP=OFF \
-DLLAMA_BUILD_EXAMPLES=ON \
-DLLAMA_BUILD_TOOLS=ON \
-DLLAMA_BUILD_TESTS=ON \
-DCMAKE_C_COMPILER_LAUNCHER=ccache \
-DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
-DCMAKE_C_COMPILER=riscv64-linux-gnu-gcc-14 \
-DCMAKE_CXX_COMPILER=riscv64-linux-gnu-g++-14 \
-DGGML_RPC=ON
cmake --build build --config Release -j $(nproc)
- name: Test
id: cmake_test
run: |
cd build
ctest -L main --verbose
ggml-ci-arm64-graviton4-kleidiai:
runs-on: ah-ubuntu_22_04-c8g_8x

166
.github/workflows/release.yml vendored
View File

@ -66,14 +66,21 @@ jobs:
id: pack_artifacts
run: |
cp LICENSE ./build/bin/
zip -r llama-${{ steps.tag.outputs.name }}-bin-macos-arm64.zip ./build/bin/*
zip -y -r llama-${{ steps.tag.outputs.name }}-bin-macos-arm64.zip ./build/bin/*
tar -czvf llama-${{ steps.tag.outputs.name }}-bin-macos-arm64.tar.gz -s ",./,llama-${{ steps.tag.outputs.name }}/," -C ./build/bin .
- name: Upload artifacts
- name: Upload artifacts (zip)
uses: actions/upload-artifact@v4
with:
path: llama-${{ steps.tag.outputs.name }}-bin-macos-arm64.zip
name: llama-bin-macos-arm64.zip
- name: Upload artifacts (tar)
uses: actions/upload-artifact@v4
with:
path: llama-${{ steps.tag.outputs.name }}-bin-macos-arm64.tar.gz
name: llama-bin-macos-arm64.tar.gz
macOS-x64:
runs-on: macos-15-intel
@ -120,14 +127,21 @@ jobs:
id: pack_artifacts
run: |
cp LICENSE ./build/bin/
zip -r llama-${{ steps.tag.outputs.name }}-bin-macos-x64.zip ./build/bin/*
zip -y -r llama-${{ steps.tag.outputs.name }}-bin-macos-x64.zip ./build/bin/*
tar -czvf llama-${{ steps.tag.outputs.name }}-bin-macos-x64.tar.gz -s ",./,llama-${{ steps.tag.outputs.name }}/," -C ./build/bin .
- name: Upload artifacts
- name: Upload artifacts (zip)
uses: actions/upload-artifact@v4
with:
path: llama-${{ steps.tag.outputs.name }}-bin-macos-x64.zip
name: llama-bin-macos-x64.zip
- name: Upload artifacts (tar)
uses: actions/upload-artifact@v4
with:
path: llama-${{ steps.tag.outputs.name }}-bin-macos-x64.tar.gz
name: llama-bin-macos-x64.tar.gz
ubuntu-22-cpu:
strategy:
matrix:
@ -182,14 +196,21 @@ jobs:
id: pack_artifacts
run: |
cp LICENSE ./build/bin/
zip -r llama-${{ steps.tag.outputs.name }}-bin-ubuntu-${{ matrix.build }}.zip ./build/bin/*
zip -y -r llama-${{ steps.tag.outputs.name }}-bin-ubuntu-${{ matrix.build }}.zip ./build/bin/*
tar -czvf llama-${{ steps.tag.outputs.name }}-bin-ubuntu-${{ matrix.build }}.tar.gz --transform "s,./,llama-${{ steps.tag.outputs.name }}/," -C ./build/bin .
- name: Upload artifacts
- name: Upload artifacts (zip)
uses: actions/upload-artifact@v4
with:
path: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-${{ matrix.build }}.zip
name: llama-bin-ubuntu-${{ matrix.build }}.zip
- name: Upload artifacts (tar)
uses: actions/upload-artifact@v4
with:
path: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-${{ matrix.build }}.tar.gz
name: llama-bin-ubuntu-${{ matrix.build }}.tar.gz
ubuntu-22-vulkan:
runs-on: ubuntu-22.04
@ -235,14 +256,21 @@ jobs:
id: pack_artifacts
run: |
cp LICENSE ./build/bin/
zip -r llama-${{ steps.tag.outputs.name }}-bin-ubuntu-vulkan-x64.zip ./build/bin/*
zip -y -r llama-${{ steps.tag.outputs.name }}-bin-ubuntu-vulkan-x64.zip ./build/bin/*
tar -czvf llama-${{ steps.tag.outputs.name }}-bin-ubuntu-vulkan-x64.tar.gz --transform "s,./,llama-${{ steps.tag.outputs.name }}/," -C ./build/bin .
- name: Upload artifacts
- name: Upload artifacts (zip)
uses: actions/upload-artifact@v4
with:
path: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-vulkan-x64.zip
name: llama-bin-ubuntu-vulkan-x64.zip
- name: Upload artifacts (tar)
uses: actions/upload-artifact@v4
with:
path: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-vulkan-x64.tar.gz
name: llama-bin-ubuntu-vulkan-x64.tar.gz
windows-cpu:
runs-on: windows-2025
@ -298,7 +326,7 @@ jobs:
run: |
Copy-Item $env:CURL_PATH\bin\libcurl-${{ matrix.arch }}.dll .\build\bin\Release\
Copy-Item "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Redist\MSVC\14.44.35112\debug_nonredist\${{ matrix.arch }}\Microsoft.VC143.OpenMP.LLVM\libomp140.${{ matrix.arch == 'x64' && 'x86_64' || 'aarch64' }}.dll" .\build\bin\Release\
7z a llama-bin-win-cpu-${{ matrix.arch }}.zip .\build\bin\Release\*
7z a -snl llama-bin-win-cpu-${{ matrix.arch }}.zip .\build\bin\Release\*
- name: Upload artifacts
uses: actions/upload-artifact@v4
@ -380,7 +408,7 @@ jobs:
- name: Pack artifacts
id: pack_artifacts
run: |
7z a llama-bin-win-${{ matrix.backend }}-${{ matrix.arch }}.zip .\build\bin\Release\${{ matrix.target }}.dll
7z a -snl llama-bin-win-${{ matrix.backend }}-${{ matrix.arch }}.zip .\build\bin\Release\${{ matrix.target }}.dll
- name: Upload artifacts
uses: actions/upload-artifact@v4
@ -393,7 +421,7 @@ jobs:
strategy:
matrix:
cuda: ['12.4']
cuda: ['12.4', '13.1']
steps:
- name: Clone
@ -434,7 +462,7 @@ jobs:
- name: Pack artifacts
id: pack_artifacts
run: |
7z a llama-bin-win-cuda-${{ matrix.cuda }}-x64.zip .\build\bin\Release\ggml-cuda.dll
7z a -snl llama-bin-win-cuda-${{ matrix.cuda }}-x64.zip .\build\bin\Release\ggml-cuda.dll
- name: Upload artifacts
uses: actions/upload-artifact@v4
@ -448,6 +476,7 @@ jobs:
$dst='.\build\bin\cudart\'
robocopy "${{env.CUDA_PATH}}\bin" $dst cudart64_*.dll cublas64_*.dll cublasLt64_*.dll
robocopy "${{env.CUDA_PATH}}\lib" $dst cudart64_*.dll cublas64_*.dll cublasLt64_*.dll
robocopy "${{env.CUDA_PATH}}\bin\x64" $dst cudart64_*.dll cublas64_*.dll cublasLt64_*.dll
7z a cudart-llama-bin-win-cuda-${{ matrix.cuda }}-x64.zip $dst\*
- name: Upload Cuda runtime
@ -517,6 +546,8 @@ jobs:
cp "${{ env.ONEAPI_ROOT }}/compiler/latest/bin/libmmd.dll" ./build/bin
cp "${{ env.ONEAPI_ROOT }}/compiler/latest/bin/libiomp5md.dll" ./build/bin
cp "${{ env.ONEAPI_ROOT }}/compiler/latest/bin/sycl-ls.exe" ./build/bin
cp "${{ env.ONEAPI_ROOT }}/compiler/latest/bin/libsycl-fallback-bfloat16.spv" ./build/bin
cp "${{ env.ONEAPI_ROOT }}/compiler/latest/bin/libsycl-native-bfloat16.spv" ./build/bin
cp "${{ env.ONEAPI_ROOT }}/dnnl/latest/bin/dnnl.dll" ./build/bin
cp "${{ env.ONEAPI_ROOT }}/tbb/latest/bin/tbb12.dll" ./build/bin
@ -526,7 +557,7 @@ jobs:
cp "${{ env.ONEAPI_ROOT }}/umf/latest/bin/umf.dll" ./build/bin
echo "cp oneAPI running time dll files to ./build/bin done"
7z a llama-bin-win-sycl-x64.zip ./build/bin/*
7z a -snl llama-bin-win-sycl-x64.zip ./build/bin/*
- name: Upload the release package
uses: actions/upload-artifact@v4
@ -632,7 +663,7 @@ jobs:
- name: Pack artifacts
id: pack_artifacts
run: |
7z a llama-bin-win-hip-${{ matrix.name }}-x64.zip .\build\bin\*
7z a -snl llama-bin-win-hip-${{ matrix.name }}-x64.zip .\build\bin\*
- name: Upload artifacts
uses: actions/upload-artifact@v4
@ -685,13 +716,21 @@ jobs:
- name: Pack artifacts
id: pack_artifacts
run: |
zip --symlinks -r llama-${{ steps.tag.outputs.name }}-xcframework.zip build-apple/llama.xcframework
zip -y -r llama-${{ steps.tag.outputs.name }}-xcframework.zip build-apple/llama.xcframework
tar -czvf llama-${{ steps.tag.outputs.name }}-xcframework.tar.gz -C build-apple llama.xcframework
- name: Upload artifacts
- name: Upload artifacts (zip)
uses: actions/upload-artifact@v4
with:
path: llama-${{ steps.tag.outputs.name }}-xcframework.zip
name: llama-${{ steps.tag.outputs.name }}-xcframework
name: llama-${{ steps.tag.outputs.name }}-xcframework.zip
- name: Upload artifacts (tar)
uses: actions/upload-artifact@v4
with:
path: llama-${{ steps.tag.outputs.name }}-xcframework.tar.gz
name: llama-${{ steps.tag.outputs.name }}-xcframework.tar.gz
openEuler-cann:
strategy:
@ -700,28 +739,54 @@ jobs:
chip_type: ['910b', '310p']
build: ['Release']
runs-on: ${{ matrix.arch == 'aarch64' && 'ubuntu-24.04-arm' || 'ubuntu-24.04' }}
container: ascendai/cann:${{ matrix.chip_type == '910b' && '8.3.rc1.alpha001-910b-openeuler22.03-py3.11' || '8.2.rc1-310p-openeuler22.03-py3.11' }}
steps:
- name: Checkout
uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Dependencies
- name: Free up disk space
uses: ggml-org/free-disk-space@v1.3.1
with:
tool-cache: true
- name: Set container image
id: cann-image
run: |
yum update -y
yum install -y git gcc gcc-c++ make cmake libcurl-devel
git config --global --add safe.directory "$GITHUB_WORKSPACE"
image="ascendai/cann:${{ matrix.chip_type == '910b' && '8.3.rc2-910b-openeuler24.03-py3.11' || '8.3.rc2-310p-openeuler24.03-py3.11' }}"
echo "image=${image}" >> "${GITHUB_OUTPUT}"
- name: Pull container image
run: docker pull "${{ steps.cann-image.outputs.image }}"
- name: Build
env:
BUILD_TYPE: ${{ matrix.build }}
SOC_TYPE: ascend${{ matrix.chip_type }}
run: |
export LD_LIBRARY_PATH=${ASCEND_TOOLKIT_HOME}/lib64:${ASCEND_TOOLKIT_HOME}/$(uname -m)-linux/devlib/:${LD_LIBRARY_PATH}
HOST_UID=$(id -u)
HOST_GID=$(id -g)
cmake -S . -B build \
-DCMAKE_BUILD_TYPE=${{ matrix.build }} \
-DGGML_CANN=on \
-DSOC_TYPE=ascend${{ matrix.chip_type }}
cmake --build build -j $(nproc)
docker run --rm \
-v "${PWD}:/workspace" \
-w /workspace \
-e SOC_TYPE=${SOC_TYPE} \
-e BUILD_TYPE=${BUILD_TYPE} \
"${{ steps.cann-image.outputs.image }}" \
bash -lc '
set -e
yum install -y --setopt=install_weak_deps=False --setopt=tsflags=nodocs git gcc gcc-c++ make cmake libcurl-devel
yum clean all && rm -rf /var/cache/yum
git config --global --add safe.directory "/workspace"
export LD_LIBRARY_PATH=${ASCEND_TOOLKIT_HOME}/lib64:${ASCEND_TOOLKIT_HOME}/$(uname -m)-linux/devlib/:${LD_LIBRARY_PATH}
cmake -S . -B build \
-DCMAKE_BUILD_TYPE=${BUILD_TYPE} \
-DGGML_CANN=on \
-DSOC_TYPE=${SOC_TYPE}
cmake --build build -j $(nproc)
chown -R '"${HOST_UID}"':'"${HOST_GID}"' /workspace/build
'
- name: Determine tag name
id: tag
@ -730,13 +795,13 @@ jobs:
- name: Pack artifacts
run: |
cp LICENSE ./build/bin/
zip -r llama-${{ steps.tag.outputs.name }}-bin-${{ matrix.chip_type }}-openEuler-${{ matrix.arch }}.zip ./build/bin/*
tar -czvf llama-${{ steps.tag.outputs.name }}-bin-${{ matrix.chip_type }}-openEuler-${{ matrix.arch }}.tar.gz --transform "s,./,llama-${{ steps.tag.outputs.name }}/," -C ./build/bin .
- name: Upload artifacts
- name: Upload artifacts (tar)
uses: actions/upload-artifact@v4
with:
path: llama-${{ steps.tag.outputs.name }}-bin-${{ matrix.chip_type }}-openEuler-${{ matrix.arch }}.zip
name: llama-bin-${{ matrix.chip_type }}-openEuler-${{ matrix.arch }}.zip
path: llama-${{ steps.tag.outputs.name }}-bin-${{ matrix.chip_type }}-openEuler-${{ matrix.arch }}.tar.gz
name: llama-bin-${{ matrix.chip_type }}-openEuler-${{ matrix.arch }}.tar.gz
release:
if: ${{ ( github.event_name == 'push' && github.ref == 'refs/heads/master' ) || github.event.inputs.create_release == 'true' }}
@ -814,6 +879,7 @@ jobs:
echo "Moving other artifacts..."
mv -v artifact/*.zip release
mv -v artifact/*.tar.gz release
- name: Create release
id: create_release
@ -822,6 +888,40 @@ jobs:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
with:
tag_name: ${{ steps.tag.outputs.name }}
body: |
> [!WARNING]
> **Release Format Update**: Linux releases will soon use .tar.gz archives instead of .zip. Please make the necessary changes to your deployment scripts.
<details open>
${{ github.event.head_commit.message }}
</details>
**macOS/iOS:**
- [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-macos-arm64.tar.gz)
- [macOS Intel (x64)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-macos-x64.tar.gz)
- [iOS XCFramework](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-xcframework.tar.gz)
**Linux:**
- [Ubuntu x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-x64.tar.gz)
- [Ubuntu x64 (Vulkan)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-vulkan-x64.tar.gz)
- [Ubuntu s390x (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-s390x.tar.gz)
**Windows:**
- [Windows x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-win-cpu-x64.zip)
- [Windows arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-win-cpu-arm64.zip)
- [Windows x64 (CUDA 12)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-win-cuda-12.4-x64.zip)
- [Windows x64 (CUDA 13)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-win-cuda-13.1-x64.zip)
- [Windows x64 (Vulkan)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-win-vulkan-x64.zip)
- [Windows x64 (SYCL)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-win-sycl-x64.zip)
- [Windows x64 (HIP)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-win-hip-radeon-x64.zip)
**openEuler:**
- [openEuler x86 (310p)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-310p-openEuler-x86.tar.gz)
- [openEuler x86 (910b)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-910b-openEuler-x86.tar.gz)
- [openEuler aarch64 (310p)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-310p-openEuler-aarch64.tar.gz)
- [openEuler aarch64 (910b)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-910b-openEuler-aarch64.tar.gz)
- name: Upload release
id: upload_release
@ -833,7 +933,7 @@ jobs:
const fs = require('fs');
const release_id = '${{ steps.create_release.outputs.id }}';
for (let file of await fs.readdirSync('./release')) {
if (path.extname(file) === '.zip') {
if (path.extname(file) === '.zip' || file.endsWith('.tar.gz')) {
console.log('uploadReleaseAsset', file);
await github.repos.uploadReleaseAsset({
owner: context.repo.owner,

1
.github/workflows/winget.yml vendored
View File

@ -9,6 +9,7 @@ jobs:
update:
name: Update Winget Package
runs-on: ubuntu-latest
if: github.repository_owner == 'ggml-org'
steps:
- name: Install cargo binstall

3
.gitignore vendored
View File

@ -54,6 +54,7 @@
/out/
/tmp/
/autogen-*.md
/common/build-info.cpp
# Deprecated
@ -134,3 +135,5 @@ poetry.toml
# IDE
/*.code-workspace
/.windsurf/
# emscripten
a.out.*

27
CMakeLists.txt
View File

@ -33,10 +33,24 @@ endif()
option(LLAMA_USE_SYSTEM_GGML "Use system libggml" OFF)
option(LLAMA_WASM_MEM64 "llama: use 64-bit memory in WASM builds" ON)
if (EMSCRIPTEN)
set(BUILD_SHARED_LIBS_DEFAULT OFF)
option(LLAMA_WASM_SINGLE_FILE "llama: embed WASM inside the generated llama.js" ON)
# Use 64-bit memory to support backend_get_memory queries
# TODO: analyze performance impact, see https://spidermonkey.dev/blog/2025/01/15/is-memory64-actually-worth-using
if (LLAMA_WASM_MEM64)
add_compile_options("-sMEMORY64=1")
add_link_options("-sMEMORY64=1")
endif()
add_link_options("-sALLOW_MEMORY_GROWTH=1")
option(LLAMA_WASM_SINGLE_FILE "llama: embed WASM inside the generated llama.js" OFF)
option(LLAMA_BUILD_HTML "llama: build HTML file" ON)
if (LLAMA_BUILD_HTML)
set(CMAKE_EXECUTABLE_SUFFIX ".html")
endif()
else()
if (MINGW)
set(BUILD_SHARED_LIBS_DEFAULT OFF)
@ -58,6 +72,12 @@ if (MSVC)
add_compile_options("$<$<COMPILE_LANGUAGE:CXX>:/bigobj>")
endif()
if (LLAMA_STANDALONE)
# enable parallel builds for msbuild
list(APPEND CMAKE_VS_GLOBALS UseMultiToolTask=true)
list(APPEND CMAKE_VS_GLOBALS EnforceProcessCountAcrossBuilds=true)
endif()
if (CMAKE_SYSTEM_NAME STREQUAL "iOS")
set(LLAMA_TOOLS_INSTALL_DEFAULT OFF)
else()
@ -179,11 +199,6 @@ if (NOT TARGET ggml AND NOT LLAMA_USE_SYSTEM_GGML)
# ... otherwise assume ggml is added by a parent CMakeLists.txt
endif()
if (MINGW)
# Target Windows 8 for PrefetchVirtualMemory
add_compile_definitions(_WIN32_WINNT=${GGML_WIN_VER})
endif()
#
# build the library
#

13
CODEOWNERS
View File

@ -7,16 +7,20 @@
/ci/ @ggerganov
/cmake/ @ggerganov
/common/CMakeLists.txt @ggerganov
/common/arg.* @ggerganov @ericcurtin
/common/arg.* @ggerganov
/common/base64.hpp.* @ggerganov
/common/build-info.* @ggerganov
/common/chat.* @pwilkin
/common/chat-peg-parser.* @aldehir
/common/common.* @ggerganov
/common/console.* @ggerganov
/common/http.* @angt
/common/llguidance.* @ggerganov
/common/log.* @ggerganov
/common/peg-parser.* @aldehir
/common/sampling.* @ggerganov
/common/speculative.* @ggerganov
/common/unicode.* @aldehir
/convert_*.py @CISC
/examples/batched.swift/ @ggerganov
/examples/batched/ @ggerganov
@ -81,14 +85,15 @@
/src/llama-vocab.* @CISC
/src/models/ @CISC
/tests/ @ggerganov
/tests/test-chat-.* @pwilkin
/tools/batched-bench/ @ggerganov
/tools/main/ @ggerganov
/tools/cli/ @ngxson
/tools/completion/ @ggerganov
/tools/mtmd/ @ngxson
/tools/perplexity/ @ggerganov
/tools/quantize/ @ggerganov
/tools/rpc/ @rgerganov
/tools/run/ @ericcurtin
/tools/server/* @ngxson @ggerganov @ericcurtin # no subdir
/tools/server/* @ngxson @ggerganov # no subdir
/tools/server/webui/ @allozaur
/tools/tokenize/ @ggerganov
/tools/tts/ @ggerganov

4
CONTRIBUTING.md
View File

@ -15,10 +15,12 @@ The project differentiates between 3 levels of contributors:
- If you modified the `ggml` source, run the `test-backend-ops` tool to check whether different backend implementations of the `ggml` operators produce consistent results (this requires access to at least two different `ggml` backends)
- If you modified a `ggml` operator or added a new one, add the corresponding test cases to `test-backend-ops`
- Create separate PRs for each feature or fix. Avoid combining unrelated changes in a single PR
- When adding support for a new model or feature, focus on **CPU support only** in the initial PR unless you have a good reason not to. Add support for other backends like CUDA in follow-up PRs
- Consider allowing write access to your branch for faster reviews, as reviewers can push commits directly
- If your PR becomes stale, don't hesitate to ping the maintainers in the comments
- If your PR becomes stale, rebase it on top of latest `master` to get maintainers attention
- Maintainers will rely on your insights and approval when making a final decision to approve and merge a PR
- Consider adding yourself to [CODEOWNERS](CODEOWNERS) to indicate your availability for reviewing related PRs
- Using AI to generate PRs is permitted. However, you must (1) explicitly disclose how AI was used and (2) conduct a thorough manual review before publishing the PR. Note that trivial tab autocompletions do not require disclosure.
# Pull requests (for maintainers)

22
README.md
View File

@ -61,7 +61,7 @@ range of hardware - locally and in the cloud.
- Plain C/C++ implementation without any dependencies
- Apple silicon is a first-class citizen - optimized via ARM NEON, Accelerate and Metal frameworks
- AVX, AVX2, AVX512 and AMX support for x86 architectures
- RVV, ZVFH, ZFH and ZICBOP support for RISC-V architectures
- RVV, ZVFH, ZFH, ZICBOP and ZIHINTPAUSE support for RISC-V architectures
- 1.5-bit, 2-bit, 3-bit, 4-bit, 5-bit, 6-bit, and 8-bit integer quantization for faster inference and reduced memory use
- Custom CUDA kernels for running LLMs on NVIDIA GPUs (support for AMD GPUs via HIP and Moore Threads GPUs via MUSA)
- Vulkan and SYCL backend support
@ -276,6 +276,7 @@ Instructions for adding support for new models: [HOWTO-add-model.md](docs/develo
| [MUSA](docs/build.md#musa) | Moore Threads GPU |
| [CUDA](docs/build.md#cuda) | Nvidia GPU |
| [HIP](docs/build.md#hip) | AMD GPU |
| [ZenDNN](docs/build.md#zendnn) | AMD CPU |
| [Vulkan](docs/build.md#vulkan) | GPU |
| [CANN](docs/build.md#cann) | Ascend NPU |
| [OpenCL](docs/backend/OPENCL.md) | Adreno GPU |
@ -312,7 +313,7 @@ The Hugging Face platform provides a variety of online tools for converting, qua
To learn more about model quantization, [read this documentation](tools/quantize/README.md)
## [`llama-cli`](tools/main)
## [`llama-cli`](tools/cli)
#### A CLI tool for accessing and experimenting with most of `llama.cpp`'s functionality.
@ -346,19 +347,6 @@ To learn more about model quantization, [read this documentation](tools/quantize
</details>
- <details>
<summary>Run simple text completion</summary>
To disable conversation mode explicitly, use `-no-cnv`
```bash
llama-cli -m model.gguf -p "I believe the meaning of life is" -n 128 -no-cnv
# I believe the meaning of life is to find your own truth and to live in accordance with it. For me, this means being true to myself and following my passions, even if they don't align with societal expectations. I think that's what I love about yoga – it's not just a physical practice, but a spiritual one too. It's about connecting with yourself, listening to your inner voice, and honoring your own unique journey.
```
</details>
- <details>
<summary>Constrain the output with a custom grammar</summary>
@ -537,7 +525,8 @@ To learn more about model quantization, [read this documentation](tools/quantize
## Other documentation
- [main (cli)](tools/main/README.md)
- [cli](tools/cli/README.md)
- [completion](tools/completion/README.md)
- [server](tools/server/README.md)
- [GBNF grammars](grammars/README.md)
@ -613,3 +602,4 @@ $ echo "source ~/.llama-completion.bash" >> ~/.bashrc
- [linenoise.cpp](./tools/run/linenoise.cpp/linenoise.cpp) - C++ library that provides readline-like line editing capabilities, used by `llama-run` - BSD 2-Clause License
- [curl](https://curl.se/) - Client-side URL transfer library, used by various tools/examples - [CURL License](https://curl.se/docs/copyright.html)
- [miniaudio.h](https://github.com/mackron/miniaudio) - Single-header audio format decoder, used by multimodal subsystem - Public domain
- [subprocess.h](https://github.com/sheredom/subprocess.h) - Single-header process launching solution for C and C++ - Public domain

2
SECURITY.md
View File

@ -65,4 +65,6 @@ However, If you have discovered a security vulnerability in this project, please
Please disclose it as a private [security advisory](https://github.com/ggml-org/llama.cpp/security/advisories/new).
Please note that using AI to identify vulnerabilities and generate reports is permitted. However, you must (1) explicitly disclose how AI was used and (2) conduct a thorough manual review before submitting the report.
A team of volunteers on a reasonable-effort basis maintains this project. As such, please give us at least 90 days to work on a fix before public exposure.

32
ci/run.sh
View File

@ -45,7 +45,7 @@ sd=`dirname $0`
cd $sd/../
SRC=`pwd`
CMAKE_EXTRA="-DLLAMA_FATAL_WARNINGS=ON -DLLAMA_CURL=ON -DGGML_SCHED_NO_REALLOC=ON"
CMAKE_EXTRA="-DLLAMA_FATAL_WARNINGS=${LLAMA_FATAL_WARNINGS:-ON} -DLLAMA_CURL=ON -DGGML_SCHED_NO_REALLOC=ON"
if [ ! -z ${GG_BUILD_METAL} ]; then
CMAKE_EXTRA="${CMAKE_EXTRA} -DGGML_METAL=ON"
@ -398,18 +398,20 @@ function gg_run_qwen3_0_6b {
./bin/llama-quantize ${model_bf16} ${model_q5_k} q5_k $(nproc)
./bin/llama-quantize ${model_bf16} ${model_q6_k} q6_k $(nproc)
(time ./bin/llama-cli -no-cnv --model ${model_f16} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-f16.log
(time ./bin/llama-cli -no-cnv --model ${model_bf16} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-bf16.log
(time ./bin/llama-cli -no-cnv --model ${model_q8_0} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q8_0.log
(time ./bin/llama-cli -no-cnv --model ${model_q4_0} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q4_0.log
(time ./bin/llama-cli -no-cnv --model ${model_q4_1} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q4_1.log
(time ./bin/llama-cli -no-cnv --model ${model_q5_0} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q5_0.log
(time ./bin/llama-cli -no-cnv --model ${model_q5_1} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q5_1.log
(time ./bin/llama-cli -no-cnv --model ${model_q2_k} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q2_k.log
(time ./bin/llama-cli -no-cnv --model ${model_q3_k} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q3_k.log
(time ./bin/llama-cli -no-cnv --model ${model_q4_k} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q4_k.log
(time ./bin/llama-cli -no-cnv --model ${model_q5_k} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q5_k.log
(time ./bin/llama-cli -no-cnv --model ${model_q6_k} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q6_k.log
(time ./bin/llama-fit-params --model ${model_f16} 2>&1 | tee -a $OUT/${ci}-fp-f16.log)
(time ./bin/llama-completion -no-cnv --model ${model_f16} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-f16.log
(time ./bin/llama-completion -no-cnv --model ${model_bf16} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-bf16.log
(time ./bin/llama-completion -no-cnv --model ${model_q8_0} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q8_0.log
(time ./bin/llama-completion -no-cnv --model ${model_q4_0} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q4_0.log
(time ./bin/llama-completion -no-cnv --model ${model_q4_1} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q4_1.log
(time ./bin/llama-completion -no-cnv --model ${model_q5_0} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q5_0.log
(time ./bin/llama-completion -no-cnv --model ${model_q5_1} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q5_1.log
(time ./bin/llama-completion -no-cnv --model ${model_q2_k} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q2_k.log
(time ./bin/llama-completion -no-cnv --model ${model_q3_k} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q3_k.log
(time ./bin/llama-completion -no-cnv --model ${model_q4_k} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q4_k.log
(time ./bin/llama-completion -no-cnv --model ${model_q5_k} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q5_k.log
(time ./bin/llama-completion -no-cnv --model ${model_q6_k} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-q6_k.log
(time ./bin/llama-perplexity --model ${model_f16} -f ${wiki_test} -ngl 99 -c 1024 -b 512 --chunks 2 ) 2>&1 | tee -a $OUT/${ci}-tg-f16.log
if [ -z ${GG_BUILD_NO_BF16} ]; then
@ -523,6 +525,8 @@ function gg_run_embd_bge_small {
./bin/llama-quantize ${model_f16} ${model_q8_0} q8_0
(time ./bin/llama-fit-params --model ${model_f16} 2>&1 | tee -a $OUT/${ci}-fp-f16.log)
(time ./bin/llama-embedding --model ${model_f16} -p "I believe the meaning of life is" -ngl 99 -c 0 --no-op-offload) 2>&1 | tee -a $OUT/${ci}-tg-f16.log
(time ./bin/llama-embedding --model ${model_q8_0} -p "I believe the meaning of life is" -ngl 99 -c 0 --no-op-offload) 2>&1 | tee -a $OUT/${ci}-tg-q8_0.log
@ -563,6 +567,8 @@ function gg_run_rerank_tiny {
model_f16="${path_models}/ggml-model-f16.gguf"
(time ./bin/llama-fit-params --model ${model_f16} 2>&1 | tee -a $OUT/${ci}-fp-f16.log)
# for this model, the SEP token is "</s>"
(time ./bin/llama-embedding --model ${model_f16} -p "what is panda?\thi\nwhat is panda?\tit's a bear\nwhat is panda?\tThe giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China." -ngl 99 -c 0 --pooling rank --embd-normalize -1 --no-op-offload --verbose-prompt) 2>&1 | tee -a $OUT/${ci}-rk-f16.log

26
cmake/build-info.cmake
View File

@ -39,26 +39,10 @@ if(Git_FOUND)
endif()
endif()
if(MSVC)
set(BUILD_COMPILER "${CMAKE_C_COMPILER_ID} ${CMAKE_C_COMPILER_VERSION}")
if (CMAKE_VS_PLATFORM_NAME)
set(BUILD_TARGET ${CMAKE_VS_PLATFORM_NAME})
else()
set(BUILD_TARGET "${CMAKE_SYSTEM_NAME} ${CMAKE_SYSTEM_PROCESSOR}")
endif()
else()
execute_process(
COMMAND ${CMAKE_C_COMPILER} --version
OUTPUT_VARIABLE OUT
OUTPUT_STRIP_TRAILING_WHITESPACE
)
string(REGEX REPLACE " *\n.*" "" OUT "${OUT}")
set(BUILD_COMPILER ${OUT})
set(BUILD_COMPILER "${CMAKE_C_COMPILER_ID} ${CMAKE_C_COMPILER_VERSION}")
execute_process(
COMMAND ${CMAKE_C_COMPILER} -dumpmachine
OUTPUT_VARIABLE OUT
OUTPUT_STRIP_TRAILING_WHITESPACE
)
set(BUILD_TARGET ${OUT})
if(CMAKE_VS_PLATFORM_NAME)
set(BUILD_TARGET ${CMAKE_VS_PLATFORM_NAME})
else()
set(BUILD_TARGET "${CMAKE_SYSTEM_NAME} ${CMAKE_SYSTEM_PROCESSOR}")
endif()

8
common/CMakeLists.txt
View File

@ -52,6 +52,8 @@ add_library(${TARGET} STATIC
chat-parser.h
chat-parser-xml-toolcall.h
chat-parser-xml-toolcall.cpp
chat-peg-parser.cpp
chat-peg-parser.h
chat.cpp
chat.h
common.cpp
@ -69,12 +71,18 @@ add_library(${TARGET} STATIC
log.h
ngram-cache.cpp
ngram-cache.h
peg-parser.cpp
peg-parser.h
preset.cpp
preset.h
regex-partial.cpp
regex-partial.h
sampling.cpp
sampling.h
speculative.cpp
speculative.h
unicode.cpp
unicode.h
)
if (BUILD_SHARED_LIBS)

604
common/arg.cpp
View File

File diff suppressed because it is too large Load Diff

45
common/arg.h
View File

@ -3,8 +3,10 @@
#include "common.h"
#include <set>
#include <map>
#include <string>
#include <vector>
#include <cstring>
//
// CLI argument parsing
@ -14,6 +16,7 @@ struct common_arg {
std::set<enum llama_example> examples = {LLAMA_EXAMPLE_COMMON};
std::set<enum llama_example> excludes = {};
std::vector<const char *> args;
std::vector<const char *> args_neg; // for negated args like --no-xxx
const char * value_hint = nullptr; // help text or example for arg value
const char * value_hint_2 = nullptr; // for second arg value
const char * env = nullptr;
@ -23,6 +26,9 @@ struct common_arg {
void (*handler_string) (common_params & params, const std::string &) = nullptr;
void (*handler_str_str)(common_params & params, const std::string &, const std::string &) = nullptr;
void (*handler_int) (common_params & params, int) = nullptr;
void (*handler_bool) (common_params & params, bool) = nullptr;
common_arg() = default;
common_arg(
const std::initializer_list<const char *> & args,
@ -44,6 +50,13 @@ struct common_arg {
void (*handler)(common_params & params)
) : args(args), help(help), handler_void(handler) {}
common_arg(
const std::initializer_list<const char *> & args,
const std::initializer_list<const char *> & args_neg,
const std::string & help,
void (*handler)(common_params & params, bool)
) : args(args), args_neg(args_neg), help(help), handler_bool(handler) {}
// support 2 values for arg
common_arg(
const std::initializer_list<const char *> & args,
@ -61,9 +74,33 @@ struct common_arg {
bool is_exclude(enum llama_example ex);
bool get_value_from_env(std::string & output) const;
bool has_value_from_env() const;
std::string to_string();
std::string to_string() const;
// for using as key in std::map
bool operator<(const common_arg& other) const {
if (args.empty() || other.args.empty()) {
return false;
}
return strcmp(args[0], other.args[0]) < 0;
}
bool operator==(const common_arg& other) const {
if (args.empty() || other.args.empty()) {
return false;
}
return strcmp(args[0], other.args[0]) == 0;
}
// get all args and env vars (including negated args/env)
std::vector<std::string> get_args() const;
std::vector<std::string> get_env() const;
};
namespace common_arg_utils {
bool is_truthy(const std::string & value);
bool is_falsey(const std::string & value);
bool is_autoy(const std::string & value);
}
struct common_params_context {
enum llama_example ex = LLAMA_EXAMPLE_COMMON;
common_params & params;
@ -76,7 +113,11 @@ struct common_params_context {
// if one argument has invalid value, it will automatically display usage of the specific argument (and not the full usage message)
bool common_params_parse(int argc, char ** argv, common_params & params, llama_example ex, void(*print_usage)(int, char **) = nullptr);
// function to be used by test-arg-parser
// parse input arguments from CLI into a map
// TODO: support repeated args in the future
bool common_params_to_map(int argc, char ** argv, llama_example ex, std::map<common_arg, std::string> & out_map);
// initialize argument parser context - used by test-arg-parser and preset
common_params_context common_params_parser_init(common_params & params, llama_example ex, void(*print_usage)(int, char **) = nullptr);
struct common_remote_params {

54
common/chat-parser-xml-toolcall.cpp
View File

@ -724,16 +724,10 @@ inline void parse_msg_with_xml_tool_calls(common_chat_msg_parser & builder, cons
if (reasoning_unclosed) {
if (auto pos = content.find(end_think); pos == std::string::npos && builder.pos() != builder.input().size()) {
unclosed_reasoning_content += content;
if (form.allow_toolcall_in_think) {
builder.move_to(tc->groups[0].begin);
if (!builder.try_consume_xml_tool_calls(form)) {
unclosed_reasoning_content += tool_call_start;
builder.move_to(tc->groups[0].end);
}
} else {
if (!(form.allow_toolcall_in_think && tc)) {
unclosed_reasoning_content += tool_call_start;
continue;
}
continue;
} else {
reasoning_unclosed = false;
std::string reasoning_content;
@ -781,8 +775,12 @@ inline void parse_msg_with_xml_tool_calls(common_chat_msg_parser & builder, cons
}
} else {
// This <tool_call> start is in thinking block, skip this tool call
auto pos = think_start + start_think.size();
unclosed_reasoning_content = content.substr(pos) + tool_call_start;
// This <tool_call> start is in thinking block
if (form.allow_toolcall_in_think) {
unclosed_reasoning_content = content.substr(think_start + start_think.size());
} else {
unclosed_reasoning_content = content.substr(think_start + start_think.size()) + tool_call_start;
}
reasoning_unclosed = true;
content.resize(think_start);
toolcall_in_think = true;
@ -805,14 +803,35 @@ inline void parse_msg_with_xml_tool_calls(common_chat_msg_parser & builder, cons
}
// remove potential partial suffix
if (content.size() > 0 && builder.pos() == builder.input().size() && unclosed_reasoning_content.empty()) {
rstrip(content);
trim_potential_partial_word(content);
rstrip(content);
if (builder.pos() == builder.input().size()) {
if (unclosed_reasoning_content.empty()) {
rstrip(content);
trim_potential_partial_word(content);
rstrip(content);
} else {
rstrip(unclosed_reasoning_content);
trim_potential_partial_word(unclosed_reasoning_content);
rstrip(unclosed_reasoning_content);
}
}
// consume unclosed_reasoning_content if allow_toolcall_in_think is set
if (form.allow_toolcall_in_think && !unclosed_reasoning_content.empty()) {
if (builder.syntax().reasoning_format != COMMON_REASONING_FORMAT_NONE && !builder.syntax().reasoning_in_content) {
builder.add_reasoning_content(unclosed_reasoning_content);
} else {
if (content.empty()) {
content = start_think + unclosed_reasoning_content;
} else {
content += "\n\n" + start_think;
content += unclosed_reasoning_content;
}
}
unclosed_reasoning_content.clear();
}
// Add content
if (content.size() != 0) {
if (!content.empty()) {
// If there are multiple content blocks
if (builder.syntax().reasoning_format != COMMON_REASONING_FORMAT_NONE && !builder.syntax().reasoning_in_content && builder.result().content.size() != 0) {
builder.add_content("\n\n");
@ -820,7 +839,7 @@ inline void parse_msg_with_xml_tool_calls(common_chat_msg_parser & builder, cons
builder.add_content(content);
}
// This <tool_call> start is in thinking block, skip this tool call
// This <tool_call> start is in thinking block and toolcall_in_think not set, skip this tool call
if (toolcall_in_think && !form.allow_toolcall_in_think) {
continue;
}
@ -829,7 +848,7 @@ inline void parse_msg_with_xml_tool_calls(common_chat_msg_parser & builder, cons
if (!tc) {
GGML_ASSERT(builder.pos() == builder.input().size());
GGML_ASSERT(unclosed_reasoning_content.empty());
GGML_ASSERT(!reasoning_unclosed);
if (!form.allow_toolcall_in_think) GGML_ASSERT(!reasoning_unclosed);
break;
}
@ -854,7 +873,6 @@ inline void parse_msg_with_xml_tool_calls(common_chat_msg_parser & builder, cons
/**
* Parse content uses reasoning and XML-Style tool call
* TODO: Note that form.allow_toolcall_in_think is not tested yet. If anyone confirms it works, this comment can be removed.
*/
void common_chat_msg_parser::consume_reasoning_with_xml_tool_calls(const struct xml_tool_call_format & form, const std::string & start_think, const std::string & end_think) {
parse_msg_with_xml_tool_calls(*this, form, start_think, end_think);

2
common/chat-parser-xml-toolcall.h
View File

@ -31,7 +31,7 @@ struct xml_tool_call_format {
std::optional<std::string> last_val_end = std::nullopt;
std::optional<std::string> last_tool_end = std::nullopt;
bool trim_raw_argval = false;
bool allow_toolcall_in_think = false; // TODO: UNTESTED!!!
bool allow_toolcall_in_think = false;
};
// make a GBNF that accept any strings except those containing any of the forbidden strings.

1009
common/chat-parser.cpp
View File

File diff suppressed because it is too large Load Diff

114
common/chat-peg-parser.cpp Normal file
View File

@ -0,0 +1,114 @@
#include "chat-peg-parser.h"
#include <nlohmann/json.hpp>
using json = nlohmann::json;
static std::string_view trim_trailing_space(std::string_view sv) {
while (!sv.empty() && std::isspace(static_cast<unsigned char>(sv.back()))) {
sv.remove_suffix(1);
}
return sv;
}
void common_chat_peg_mapper::from_ast(const common_peg_ast_arena & arena, const common_peg_parse_result & result) {
arena.visit(result, [this](const common_peg_ast_node & node) {
map(node);
});
}
void common_chat_peg_mapper::map(const common_peg_ast_node & node) {
bool is_reasoning = node.tag == common_chat_peg_builder::REASONING;
bool is_content = node.tag == common_chat_peg_builder::CONTENT;
if (is_reasoning) {
result.reasoning_content = std::string(trim_trailing_space(node.text));
}
if (is_content) {
result.content = std::string(trim_trailing_space(node.text));
}
}
void common_chat_peg_native_mapper::map(const common_peg_ast_node & node) {
common_chat_peg_mapper::map(node);
bool is_tool_open = node.tag == common_chat_peg_native_builder::TOOL_OPEN;
bool is_tool_name = node.tag == common_chat_peg_native_builder::TOOL_NAME;
bool is_tool_id = node.tag == common_chat_peg_native_builder::TOOL_ID;
bool is_tool_args = node.tag == common_chat_peg_native_builder::TOOL_ARGS;
if (is_tool_open) {
result.tool_calls.emplace_back();
current_tool = &result.tool_calls.back();
}
if (is_tool_id && current_tool) {
current_tool->id = std::string(trim_trailing_space(node.text));
}
if (is_tool_name && current_tool) {
current_tool->name = std::string(trim_trailing_space(node.text));
}
if (is_tool_args && current_tool) {
current_tool->arguments = std::string(trim_trailing_space(node.text));
}
}
void common_chat_peg_constructed_mapper::map(const common_peg_ast_node & node) {
common_chat_peg_mapper::map(node);
bool is_tool_open = node.tag == common_chat_peg_constructed_builder::TOOL_OPEN;
bool is_tool_name = node.tag == common_chat_peg_constructed_builder::TOOL_NAME;
bool is_tool_close = node.tag == common_chat_peg_constructed_builder::TOOL_CLOSE;
bool is_arg_open = node.tag == common_chat_peg_constructed_builder::TOOL_ARG_OPEN;
bool is_arg_close = node.tag == common_chat_peg_constructed_builder::TOOL_ARG_CLOSE;
bool is_arg_name = node.tag == common_chat_peg_constructed_builder::TOOL_ARG_NAME;
bool is_arg_string = node.tag == common_chat_peg_constructed_builder::TOOL_ARG_STRING_VALUE;
bool is_arg_json = node.tag == common_chat_peg_constructed_builder::TOOL_ARG_JSON_VALUE;
if (is_tool_open) {
result.tool_calls.emplace_back();
current_tool = &result.tool_calls.back();
arg_count = 0;
}
if (is_tool_name) {
current_tool->name = std::string(node.text);
current_tool->arguments = "{";
}
if (is_arg_open) {
needs_closing_quote = false;
}
if (is_arg_name && current_tool) {
if (arg_count > 0) {
current_tool->arguments += ",";
}
current_tool->arguments += json(trim_trailing_space(node.text)).dump() + ":";
++arg_count;
}
if (is_arg_string && current_tool) {
// Serialize to JSON, but exclude the end quote
std::string dumped = json(node.text).dump();
current_tool->arguments += dumped.substr(0, dumped.size() - 1);
needs_closing_quote = true;
}
if (is_arg_close && current_tool) {
if (needs_closing_quote) {
current_tool->arguments += "\"";
}
}
if (is_arg_json && current_tool) {
current_tool->arguments += std::string(trim_trailing_space(node.text));
}
if (is_tool_close && current_tool) {
current_tool->arguments += "}";
}
}

105
common/chat-peg-parser.h Normal file
View File

@ -0,0 +1,105 @@
#pragma once
#include "chat.h"
#include "peg-parser.h"
class common_chat_peg_builder : public common_peg_parser_builder {
public:
static constexpr const char * REASONING_BLOCK = "reasoning-block";
static constexpr const char * REASONING = "reasoning";
static constexpr const char * CONTENT = "content";
common_peg_parser reasoning_block(const common_peg_parser & p) { return tag(REASONING_BLOCK, p); }
common_peg_parser reasoning(const common_peg_parser & p) { return tag(REASONING, p); }
common_peg_parser content(const common_peg_parser & p) { return tag(CONTENT, p); }
};
inline common_peg_arena build_chat_peg_parser(const std::function<common_peg_parser(common_chat_peg_builder & builder)> & fn) {
common_chat_peg_builder builder;
builder.set_root(fn(builder));
return builder.build();
}
class common_chat_peg_mapper {
public:
common_chat_msg & result;
common_chat_peg_mapper(common_chat_msg & msg) : result(msg) {}
virtual void from_ast(const common_peg_ast_arena & arena, const common_peg_parse_result & result);
virtual void map(const common_peg_ast_node & node);
};
class common_chat_peg_native_builder : public common_chat_peg_builder {
public:
static constexpr const char * TOOL = "tool";
static constexpr const char * TOOL_OPEN = "tool-open";
static constexpr const char * TOOL_CLOSE = "tool-close";
static constexpr const char * TOOL_ID = "tool-id";
static constexpr const char * TOOL_NAME = "tool-name";
static constexpr const char * TOOL_ARGS = "tool-args";
common_peg_parser tool(const common_peg_parser & p) { return tag(TOOL, p); }
common_peg_parser tool_open(const common_peg_parser & p) { return atomic(tag(TOOL_OPEN, p)); }
common_peg_parser tool_close(const common_peg_parser & p) { return atomic(tag(TOOL_CLOSE, p)); }
common_peg_parser tool_id(const common_peg_parser & p) { return atomic(tag(TOOL_ID, p)); }
common_peg_parser tool_name(const common_peg_parser & p) { return atomic(tag(TOOL_NAME, p)); }
common_peg_parser tool_args(const common_peg_parser & p) { return tag(TOOL_ARGS, p); }
};
class common_chat_peg_native_mapper : public common_chat_peg_mapper {
common_chat_tool_call * current_tool;
public:
common_chat_peg_native_mapper(common_chat_msg & msg) : common_chat_peg_mapper(msg) {}
void map(const common_peg_ast_node & node) override;
};
inline common_peg_arena build_chat_peg_native_parser(const std::function<common_peg_parser(common_chat_peg_native_builder & builder)> & fn) {
common_chat_peg_native_builder builder;
builder.set_root(fn(builder));
return builder.build();
}
class common_chat_peg_constructed_builder : public common_chat_peg_builder {
public:
static constexpr const char * TOOL = "tool";
static constexpr const char * TOOL_OPEN = "tool-open";
static constexpr const char * TOOL_CLOSE = "tool-close";
static constexpr const char * TOOL_NAME = "tool-name";
static constexpr const char * TOOL_ARG = "tool-arg";
static constexpr const char * TOOL_ARG_OPEN = "tool-arg-open";
static constexpr const char * TOOL_ARG_CLOSE = "tool-arg-close";
static constexpr const char * TOOL_ARG_NAME = "tool-arg-name";
static constexpr const char * TOOL_ARG_STRING_VALUE = "tool-arg-string-value";
static constexpr const char * TOOL_ARG_JSON_VALUE = "tool-arg-json-value";
common_peg_parser tool(const common_peg_parser & p) { return tag(TOOL, p); }
common_peg_parser tool_open(const common_peg_parser & p) { return atomic(tag(TOOL_OPEN, p)); }
common_peg_parser tool_close(const common_peg_parser & p) { return atomic(tag(TOOL_CLOSE, p)); }
common_peg_parser tool_name(const common_peg_parser & p) { return atomic(tag(TOOL_NAME, p)); }
common_peg_parser tool_arg(const common_peg_parser & p) { return tag(TOOL_ARG, p); }
common_peg_parser tool_arg_open(const common_peg_parser & p) { return atomic(tag(TOOL_ARG_OPEN, p)); }
common_peg_parser tool_arg_close(const common_peg_parser & p) { return atomic(tag(TOOL_ARG_CLOSE, p)); }
common_peg_parser tool_arg_name(const common_peg_parser & p) { return atomic(tag(TOOL_ARG_NAME, p)); }
common_peg_parser tool_arg_string_value(const common_peg_parser & p) { return tag(TOOL_ARG_STRING_VALUE, p); }
common_peg_parser tool_arg_json_value(const common_peg_parser & p) { return tag(TOOL_ARG_JSON_VALUE, p); }
};
class common_chat_peg_constructed_mapper : public common_chat_peg_mapper {
common_chat_tool_call * current_tool;
int arg_count = 0;
bool needs_closing_quote = false;
public:
common_chat_peg_constructed_mapper(common_chat_msg & msg) : common_chat_peg_mapper(msg) {}
void map(const common_peg_ast_node & node) override;
};
inline common_peg_arena build_chat_peg_constructed_parser(const std::function<common_peg_parser(common_chat_peg_constructed_builder & builder)> & fn) {
common_chat_peg_constructed_builder builder;
builder.set_root(fn(builder));
return builder.build();
}

1159
common/chat.cpp
View File

File diff suppressed because it is too large Load Diff

11
common/chat.h
View File

@ -3,6 +3,7 @@
#pragma once
#include "common.h"
#include "peg-parser.h"
#include <functional>
#include <chrono>
#include <string>
@ -76,7 +77,7 @@ struct common_chat_msg_diff {
size_t tool_call_index = std::string::npos;
common_chat_tool_call tool_call_delta;
static std::vector<common_chat_msg_diff> compute_diffs(const common_chat_msg & previous_msg, const common_chat_msg & new_msg);
static std::vector<common_chat_msg_diff> compute_diffs(const common_chat_msg & msg_prv, const common_chat_msg & msg_new);
bool operator==(const common_chat_msg_diff & other) const {
return content_delta == other.content_delta
@ -124,6 +125,11 @@ enum common_chat_format {
COMMON_CHAT_FORMAT_APRIEL_1_5,
COMMON_CHAT_FORMAT_XIAOMI_MIMO,
// These are intended to be parsed by the PEG parser
COMMON_CHAT_FORMAT_PEG_SIMPLE,
COMMON_CHAT_FORMAT_PEG_NATIVE,
COMMON_CHAT_FORMAT_PEG_CONSTRUCTED,
COMMON_CHAT_FORMAT_COUNT, // Not a format, just the # formats
};
@ -154,6 +160,7 @@ struct common_chat_params {
std::vector<common_grammar_trigger> grammar_triggers;
std::vector<std::string> preserved_tokens;
std::vector<std::string> additional_stops;
std::string parser;
};
struct common_chat_syntax {
@ -163,6 +170,7 @@ struct common_chat_syntax {
bool reasoning_in_content = false;
bool thinking_forced_open = false;
bool parse_tool_calls = true;
common_peg_arena parser = {};
};
// Check if the template supplied via "--chat-template" is supported or not. Returns true if it's valid
@ -206,6 +214,7 @@ const char* common_chat_format_name(common_chat_format format);
const char* common_reasoning_format_name(common_reasoning_format format);
common_reasoning_format common_reasoning_format_from_name(const std::string & format);
common_chat_msg common_chat_parse(const std::string & input, bool is_partial, const common_chat_syntax & syntax);
common_chat_msg common_chat_peg_parse(const common_peg_arena & parser, const std::string & input, bool is_partial, const common_chat_syntax & syntax);
common_chat_tool_choice common_chat_tool_choice_parse_oaicompat(const std::string & tool_choice);

274
common/common.cpp
View File

@ -694,7 +694,7 @@ bool string_parse_kv_override(const char * data, std::vector<llama_model_kv_over
// Validate if a filename is safe to use
// To validate a full path, split the path by the OS-specific path separator, and validate each part with this function
bool fs_validate_filename(const std::string & filename) {
bool fs_validate_filename(const std::string & filename, bool allow_subdirs) {
if (!filename.length()) {
// Empty filename invalid
return false;
@ -754,10 +754,14 @@ bool fs_validate_filename(const std::string & filename) {
|| (c >= 0xD800 && c <= 0xDFFF) // UTF-16 surrogate pairs
|| c == 0xFFFD // Replacement Character (UTF-8)
|| c == 0xFEFF // Byte Order Mark (BOM)
|| c == '/' || c == '\\' || c == ':' || c == '*' // Illegal characters
|| c == ':' || c == '*' // Illegal characters
|| c == '?' || c == '"' || c == '<' || c == '>' || c == '|') {
return false;
}
if (!allow_subdirs && (c == '/' || c == '\\')) {
// Subdirectories not allowed, reject path separators
return false;
}
}
// Reject any leading or trailing ' ', or any trailing '.', these are stripped on Windows and will cause a different filename
@ -782,11 +786,29 @@ bool fs_validate_filename(const std::string & filename) {
#include <iostream>
#ifdef _WIN32
static std::wstring utf8_to_wstring(const std::string & str) {
if (str.empty()) {
return std::wstring();
}
int size = MultiByteToWideChar(CP_UTF8, 0, str.c_str(), (int)str.size(), NULL, 0);
if (size <= 0) {
return std::wstring();
}
std::wstring wstr(size, 0);
MultiByteToWideChar(CP_UTF8, 0, str.c_str(), (int)str.size(), &wstr[0], size);
return wstr;
}
#endif
// returns true if successful, false otherwise
bool fs_create_directory_with_parents(const std::string & path) {
#ifdef _WIN32
std::wstring_convert<std::codecvt_utf8<wchar_t>> converter;
std::wstring wpath = converter.from_bytes(path);
std::wstring wpath = utf8_to_wstring(path);
// if the path already exists, check whether it's a directory
const DWORD attributes = GetFileAttributesW(wpath.c_str());
@ -859,6 +881,11 @@ bool fs_create_directory_with_parents(const std::string & path) {
#endif // _WIN32
}
bool fs_is_directory(const std::string & path) {
std::filesystem::path dir(path);
return std::filesystem::exists(dir) && std::filesystem::is_directory(dir);
}
std::string fs_get_cache_directory() {
std::string cache_directory = "";
auto ensure_trailing_slash = [](std::string p) {
@ -893,6 +920,8 @@ std::string fs_get_cache_directory() {
cache_directory = std::getenv("HOME") + std::string("/Library/Caches/");
#elif defined(_WIN32)
cache_directory = std::getenv("LOCALAPPDATA");
#elif defined(__EMSCRIPTEN__)
GGML_ABORT("not implemented on this platform");
#else
# error Unknown architecture
#endif
@ -912,7 +941,7 @@ std::string fs_get_cache_file(const std::string & filename) {
return cache_directory + filename;
}
std::vector<common_file_info> fs_list_files(const std::string & path) {
std::vector<common_file_info> fs_list(const std::string & path, bool include_directories) {
std::vector<common_file_info> files;
if (path.empty()) return files;
@ -927,14 +956,22 @@ std::vector<common_file_info> fs_list_files(const std::string & path) {
const auto & p = entry.path();
if (std::filesystem::is_regular_file(p)) {
common_file_info info;
info.path = p.string();
info.name = p.filename().string();
info.path = p.string();
info.name = p.filename().string();
info.is_dir = false;
try {
info.size = static_cast<size_t>(std::filesystem::file_size(p));
} catch (const std::filesystem::filesystem_error &) {
info.size = 0;
}
files.push_back(std::move(info));
} else if (include_directories && std::filesystem::is_directory(p)) {
common_file_info info;
info.path = p.string();
info.name = p.filename().string();
info.size = 0; // Directories have no size
info.is_dir = true;
files.push_back(std::move(info));
}
} catch (const std::filesystem::filesystem_error &) {
// skip entries we cannot inspect
@ -945,36 +982,71 @@ std::vector<common_file_info> fs_list_files(const std::string & path) {
return files;
}
//
// TTY utils
//
bool tty_can_use_colors() {
// Check NO_COLOR environment variable (https://no-color.org/)
if (const char * no_color = std::getenv("NO_COLOR")) {
if (no_color[0] != '\0') {
return false;
}
}
// Check TERM environment variable
if (const char * term = std::getenv("TERM")) {
if (std::strcmp(term, "dumb") == 0) {
return false;
}
}
// Check if stdout and stderr are connected to a terminal
// We check both because log messages can go to either
bool stdout_is_tty = isatty(fileno(stdout));
bool stderr_is_tty = isatty(fileno(stderr));
return stdout_is_tty || stderr_is_tty;
}
//
// Model utils
//
static inline void common_init_sampler_from_model(
// TODO: move to common/sampling
static void common_init_sampler_from_model(
const llama_model * model,
common_params_sampling & sparams) {
const uint64_t config = sparams.user_sampling_config;
auto get_int32 = [&](const char * key, int32_t & dst, uint64_t user_config) {
if (config & user_config) return;
if (config & user_config) {
return;
}
char buf[64] = {0};
if (llama_model_meta_val_str(model, key, buf, sizeof(buf)) > 0) {
char * end = nullptr;
int32_t v = strtol(buf, &end, 10);
if (end && end != buf) dst = v;
if (end && end != buf) {
dst = v;
}
}
};
auto get_float = [&](const char * key, float & dst, uint64_t user_config) {
if (config & user_config) return;
if (config & user_config) {
return;
}
char buf[128] = {0};
if (llama_model_meta_val_str(model, key, buf, sizeof(buf)) > 0) {
char * end = nullptr;
float v = strtof(buf, &end);
if (end && end != buf) dst = v;
if (end && end != buf) {
dst = v;
}
}
};
@ -1002,31 +1074,125 @@ static inline void common_init_sampler_from_model(
get_float(llama_model_meta_key_str(LLAMA_MODEL_META_KEY_SAMPLING_MIROSTAT_ETA), sparams.mirostat_eta, common_params_sampling_config::COMMON_PARAMS_SAMPLING_CONFIG_MIROSTAT_ETA);
}
struct common_init_result common_init_from_params(common_params & params) {
common_init_result iparams;
struct common_init_result::impl {
impl() = default;
~impl() = default;
llama_model_ptr model;
llama_context_ptr context;
std::vector<llama_adapter_lora_ptr> lora;
std::vector<common_sampler_ptr> samplers;
};
common_init_result::common_init_result(common_params & params) :
pimpl(new impl{}) {
auto mparams = common_model_params_to_llama(params);
auto cparams = common_context_params_to_llama(params);
if (params.fit_params) {
LOG_INF("%s: fitting params to device memory, to report bugs during this step use -fit off (or --verbose if you can't)\n", __func__);
llama_params_fit(params.model.path.c_str(), &mparams, &cparams,
params.tensor_split, params.tensor_buft_overrides.data(), params.fit_params_target, params.fit_params_min_ctx,
params.verbosity >= 4 ? GGML_LOG_LEVEL_DEBUG : GGML_LOG_LEVEL_ERROR);
}
llama_model * model = llama_model_load_from_file(params.model.path.c_str(), mparams);
if (model == NULL) {
LOG_ERR("%s: failed to load model '%s', try reducing --n-gpu-layers if you're running out of VRAM\n",
__func__, params.model.path.c_str());
return iparams;
return;
}
common_init_sampler_from_model(model, params.sampling);
pimpl->model.reset(model);
const llama_vocab * vocab = llama_model_get_vocab(model);
auto cparams = common_context_params_to_llama(params);
// updates params.sampling
// TODO: fix naming
common_init_sampler_from_model(model, params.sampling);
if (params.sampling.ignore_eos && llama_vocab_eos(vocab) == LLAMA_TOKEN_NULL) {
LOG_WRN("%s: warning: vocab does not have an EOS token, ignoring --ignore-eos\n", __func__);
params.sampling.ignore_eos = false;
}
// initialize once
for (llama_token i = 0; i < llama_vocab_n_tokens(vocab); i++) {
if (llama_vocab_is_eog(vocab, i)) {
LOG_INF("%s: added %s logit bias = %f\n", __func__, common_token_to_piece(vocab, i).c_str(), -INFINITY);
params.sampling.logit_bias_eog.push_back({i, -INFINITY});
}
}
if (params.sampling.ignore_eos) {
// add EOG biases to the active set of logit biases
params.sampling.logit_bias.insert(
params.sampling.logit_bias.end(),
params.sampling.logit_bias_eog.begin(), params.sampling.logit_bias_eog.end());
}
//if (params.sampling.penalty_last_n == -1) {
// LOG_INF("%s: setting penalty_last_n to ctx_size = %d\n", __func__, llama_n_ctx(lctx));
// params.sampling.penalty_last_n = llama_n_ctx(lctx);
//}
//if (params.sampling.dry_penalty_last_n == -1) {
// LOG_INF("%s: setting dry_penalty_last_n to ctx_size = %d\n", __func__, llama_n_ctx(lctx));
// params.sampling.dry_penalty_last_n = llama_n_ctx(lctx);
//}
pimpl->samplers.resize(cparams.n_seq_max);
for (int i = 0; i < (int) cparams.n_seq_max; ++i) {
pimpl->samplers[i].reset(common_sampler_init(model, params.sampling));
}
llama_context * lctx = llama_init_from_model(model, cparams);
if (lctx == NULL) {
LOG_ERR("%s: failed to create context with model '%s', try reducing --n-gpu-layers if you're running out of VRAM\n",
__func__, params.model.path.c_str());
llama_model_free(model);
return iparams;
LOG_ERR("%s: failed to create context with model '%s'\n", __func__, params.model.path.c_str());
return;
}
pimpl->context.reset(lctx);
}
llama_model * common_init_result::model() {
return pimpl->model.get();
}
llama_context * common_init_result::context() {
return pimpl->context.get();
}
common_sampler * common_init_result::sampler(llama_seq_id seq_id) {
return pimpl->samplers[seq_id].get();
}
std::vector<llama_adapter_lora_ptr> & common_init_result::lora() {
return pimpl->lora;
}
void common_init_result::free_context() {
pimpl->context.reset();
}
common_init_result_ptr common_init_from_params(common_params & params) {
common_init_result_ptr res(new common_init_result(params));
llama_model * model = res->model();
if (model == NULL) {
LOG_ERR("%s: failed to load model '%s'\n", __func__, params.model.path.c_str());
return res;
}
llama_context * lctx = res->context();
if (lctx == NULL) {
LOG_ERR("%s: failed to create context with model '%s'\n", __func__, params.model.path.c_str());
return res;
}
const llama_vocab * vocab = llama_model_get_vocab(model);
if (params.ctx_shift && !llama_memory_can_shift(llama_get_memory(lctx))) {
LOG_WRN("%s: KV cache shifting is not supported for this context, disabling KV cache shifting\n", __func__);
params.ctx_shift = false;
@ -1038,10 +1204,7 @@ struct common_init_result common_init_from_params(common_params & params) {
const auto cvec = common_control_vector_load(params.control_vectors);
if (cvec.n_embd == -1) {
llama_free(lctx);
llama_model_free(model);
return iparams;
return res;
}
int err = llama_apply_adapter_cvec(
@ -1052,10 +1215,7 @@ struct common_init_result common_init_from_params(common_params & params) {
params.control_vector_layer_start,
params.control_vector_layer_end);
if (err) {
llama_free(lctx);
llama_model_free(model);
return iparams;
return res;
}
}
@ -1079,10 +1239,7 @@ struct common_init_result common_init_from_params(common_params & params) {
}
if (!ok) {
llama_free(lctx);
llama_model_free(model);
return iparams;
return res;
}
}
@ -1092,9 +1249,7 @@ struct common_init_result common_init_from_params(common_params & params) {
lora.reset(llama_adapter_lora_init(model, la.path.c_str()));
if (lora == nullptr) {
LOG_ERR("%s: failed to apply lora adapter '%s'\n", __func__, la.path.c_str());
llama_free(lctx);
llama_model_free(model);
return iparams;
return res;
}
char buf[1024];
@ -1103,43 +1258,13 @@ struct common_init_result common_init_from_params(common_params & params) {
la.task_name = buf;
llama_adapter_meta_val_str(la.ptr, "adapter.lora.prompt_prefix", buf, sizeof(buf));
la.prompt_prefix = buf;
iparams.lora.emplace_back(std::move(lora)); // copy to list of loaded adapters
res->lora().emplace_back(std::move(lora)); // copy to list of loaded adapters
}
if (!params.lora_init_without_apply) {
common_set_adapter_lora(lctx, params.lora_adapters);
}
if (params.sampling.ignore_eos && llama_vocab_eos(vocab) == LLAMA_TOKEN_NULL) {
LOG_WRN("%s: warning: vocab does not have an EOS token, ignoring --ignore-eos\n", __func__);
params.sampling.ignore_eos = false;
}
// initialize once
for (llama_token i = 0; i < llama_vocab_n_tokens(vocab); i++) {
if (llama_vocab_is_eog(vocab, i)) {
LOG_INF("%s: added %s logit bias = %f\n", __func__, common_token_to_piece(lctx, i).c_str(), -INFINITY);
params.sampling.logit_bias_eog.push_back({i, -INFINITY});
}
}
if (params.sampling.ignore_eos) {
// add EOG biases to the active set of logit biases
params.sampling.logit_bias.insert(
params.sampling.logit_bias.end(),
params.sampling.logit_bias_eog.begin(), params.sampling.logit_bias_eog.end());
}
if (params.sampling.penalty_last_n == -1) {
LOG_INF("%s: setting penalty_last_n to ctx_size = %d\n", __func__, llama_n_ctx(lctx));
params.sampling.penalty_last_n = llama_n_ctx(lctx);
}
if (params.sampling.dry_penalty_last_n == -1) {
LOG_INF("%s: setting dry_penalty_last_n to ctx_size = %d\n", __func__, llama_n_ctx(lctx));
params.sampling.dry_penalty_last_n = llama_n_ctx(lctx);
}
if (params.warmup) {
LOG_WRN("%s: warming up the model with an empty run - please wait ... (--no-warmup to disable)\n", __func__);
@ -1178,12 +1303,11 @@ struct common_init_result common_init_from_params(common_params & params) {
llama_set_warmup(lctx, false);
}
iparams.model.reset(model);
iparams.context.reset(lctx);
return iparams;
return res;
}
common_init_result::~common_init_result() = default;
std::string get_model_endpoint() {
const char * model_endpoint_env = getenv("MODEL_ENDPOINT");
// We still respect the use of environment-variable "HF_ENDPOINT" for backward-compatibility.
@ -1192,7 +1316,9 @@ std::string get_model_endpoint() {
std::string model_endpoint = "https://huggingface.co/";
if (endpoint_env) {
model_endpoint = endpoint_env;
if (model_endpoint.back() != '/') model_endpoint += '/';
if (model_endpoint.back() != '/') {
model_endpoint += '/';
}
}
return model_endpoint;
}

80
common/common.h
View File

@ -12,6 +12,10 @@
#include <vector>
#include <map>
#if defined(_WIN32) && !defined(_WIN32_WINNT)
#define _WIN32_WINNT 0x0A00
#endif
#ifdef _WIN32
#define DIRECTORY_SEPARATOR '\\'
#else
@ -26,8 +30,6 @@
fprintf(stderr, "%s: built with %s for %s\n", __func__, LLAMA_COMPILER, LLAMA_BUILD_TARGET); \
} while(0)
#define DEFAULT_MODEL_PATH "models/7B/ggml-model-f16.gguf"
struct common_time_meas {
common_time_meas(int64_t & t_acc, bool disable = false);
~common_time_meas();
@ -80,7 +82,8 @@ int32_t cpu_get_num_math();
enum llama_example {
LLAMA_EXAMPLE_COMMON,
LLAMA_EXAMPLE_SPECULATIVE,
LLAMA_EXAMPLE_MAIN,
LLAMA_EXAMPLE_COMPLETION,
LLAMA_EXAMPLE_CLI,
LLAMA_EXAMPLE_EMBEDDING,
LLAMA_EXAMPLE_PERPLEXITY,
LLAMA_EXAMPLE_RETRIEVAL,
@ -96,6 +99,7 @@ enum llama_example {
LLAMA_EXAMPLE_TTS,
LLAMA_EXAMPLE_DIFFUSION,
LLAMA_EXAMPLE_FINETUNE,
LLAMA_EXAMPLE_FIT_PARAMS,
LLAMA_EXAMPLE_COUNT,
};
@ -192,7 +196,6 @@ struct common_params_sampling {
std::vector<std::string> dry_sequence_breakers = {"\n", ":", "\"", "*"}; // default sequence breakers for DRY
std::vector<enum common_sampler_type> samplers = {
COMMON_SAMPLER_TYPE_PENALTIES,
COMMON_SAMPLER_TYPE_DRY,
@ -213,6 +216,10 @@ struct common_params_sampling {
std::vector<llama_logit_bias> logit_bias; // logit biases to apply
std::vector<llama_logit_bias> logit_bias_eog; // pre-calculated logit biases for EOG tokens
bool has_logit_bias() const {
return !logit_bias.empty();
}
// print the parameters into a string
std::string print() const;
};
@ -223,6 +230,7 @@ struct common_params_model {
std::string hf_repo = ""; // HF repo // NOLINT
std::string hf_file = ""; // HF file // NOLINT
std::string docker_repo = ""; // Docker repo // NOLINT
std::string name = ""; // in format <user>/<model>[:<tag>] (tag is optional) // NOLINT
};
struct common_params_speculative {
@ -299,8 +307,8 @@ struct lr_opt {
struct ggml_opt_optimizer_params common_opt_lr_pars(void * userdata);
struct common_params {
int32_t n_predict = -1; // new tokens to predict
int32_t n_ctx = 4096; // context size
int32_t n_predict = -1; // max. number of new tokens to predict, -1 == no limit
int32_t n_ctx = 0; // context size, 0 == context the model was trained with
int32_t n_batch = 2048; // logical batch size for prompt processing (must be >=32 to use BLAS)
int32_t n_ubatch = 512; // physical batch size for prompt processing (must be >=32 to use BLAS)
int32_t n_keep = 0; // number of tokens to keep from initial prompt
@ -321,9 +329,12 @@ struct common_params {
// offload params
std::vector<ggml_backend_dev_t> devices; // devices to use for offloading
int32_t n_gpu_layers = -1; // number of layers to store in VRAM (-1 - use default)
int32_t main_gpu = 0; // the GPU that is used for scratch and small tensors
float tensor_split[128] = {0}; // how split tensors should be distributed across GPUs
int32_t n_gpu_layers = -1; // number of layers to store in VRAM (-1 - use default)
int32_t main_gpu = 0; // the GPU that is used for scratch and small tensors
float tensor_split[128] = {0}; // how split tensors should be distributed across GPUs
bool fit_params = true; // whether to fit unset model/context parameters to free device memory
size_t fit_params_target = 1024 * 1024*1024; // margin per device in bytes for fitting parameters to free memory
int32_t fit_params_min_ctx = 4096; // minimum context size to set when trying to reduce memory use
enum llama_split_mode split_mode = LLAMA_SPLIT_MODE_LAYER; // how to split the model across GPUs
@ -369,7 +380,7 @@ struct common_params {
std::vector<common_control_vector_load_info> control_vectors; // control vector with user defined scale
int32_t verbosity = 0;
int32_t verbosity = 3; // LOG_LEVEL_INFO
int32_t control_vector_layer_start = -1; // layer range for control vector
int32_t control_vector_layer_end = -1; // layer range for control vector
bool offline = false;
@ -403,6 +414,7 @@ struct common_params {
bool simple_io = false; // improves compatibility with subprocesses and limited consoles
bool cont_batching = true; // insert new sequences for decoding on-the-fly
bool no_perf = false; // disable performance metrics
bool show_timings = true; // show timing information on CLI
bool ctx_shift = false; // context shift on infinite text generation
bool swa_full = false; // use full-size SWA cache (https://github.com/ggml-org/llama.cpp/pull/13194#issuecomment-2868343055)
bool kv_unified = false; // enable unified KV cache
@ -459,7 +471,7 @@ struct common_params {
std::string public_path = ""; // NOLINT
std::string api_prefix = ""; // NOLINT
std::string chat_template = ""; // NOLINT
bool use_jinja = false; // NOLINT
bool use_jinja = true; // NOLINT
bool enable_chat_template = true;
common_reasoning_format reasoning_format = COMMON_REASONING_FORMAT_DEEPSEEK;
int reasoning_budget = -1;
@ -478,9 +490,16 @@ struct common_params {
bool endpoint_props = false; // only control POST requests, not GET
bool endpoint_metrics = false;
// router server configs
std::string models_dir = ""; // directory containing models for the router server
std::string models_preset = ""; // directory containing model presets for the router server
int models_max = 4; // maximum number of models to load simultaneously
bool models_autoload = true; // automatically load models when requested via the router server
bool log_json = false;
std::string slot_save_path;
std::string media_path; // path to directory for loading media files
float slot_prompt_similarity = 0.1f;
@ -631,8 +650,9 @@ std::string string_from(const struct llama_context * ctx, const struct llama_bat
// Filesystem utils
//
bool fs_validate_filename(const std::string & filename);
bool fs_validate_filename(const std::string & filename, bool allow_subdirs = false);
bool fs_create_directory_with_parents(const std::string & path);
bool fs_is_directory(const std::string & path);
std::string fs_get_cache_directory();
std::string fs_get_cache_file(const std::string & filename);
@ -641,22 +661,44 @@ struct common_file_info {
std::string path;
std::string name;
size_t size = 0; // in bytes
bool is_dir = false;
};
std::vector<common_file_info> fs_list_files(const std::string & path);
std::vector<common_file_info> fs_list(const std::string & path, bool include_directories);
//
// TTY utils
//
// Auto-detect if colors can be enabled based on terminal and environment
bool tty_can_use_colors();
//
// Model utils
//
// note: defines object's lifetime
struct common_init_result {
llama_model_ptr model;
llama_context_ptr context;
struct common_sampler;
std::vector<llama_adapter_lora_ptr> lora;
// note: defines the model, context, samplers, ets. lifetimes
struct common_init_result {
common_init_result(common_params & params);
~common_init_result();
llama_model * model();
llama_context * context();
common_sampler * sampler(llama_seq_id seq_id);
std::vector<llama_adapter_lora_ptr> & lora();
void free_context();
private:
struct impl;
std::unique_ptr<impl> pimpl;
};
struct common_init_result common_init_from_params(common_params & params);
using common_init_result_ptr = std::unique_ptr<common_init_result>;
common_init_result_ptr common_init_from_params(common_params & params);
struct llama_model_params common_model_params_to_llama ( common_params & params);
struct llama_context_params common_context_params_to_llama(const common_params & params);

729
common/console.cpp
View File

@ -1,6 +1,16 @@
#include "console.h"
#include "log.h"
#include <vector>
#include <iostream>
#include <cassert>
#include <cstddef>
#include <cctype>
#include <cwctype>
#include <cstdint>
#include <condition_variable>
#include <mutex>
#include <thread>
#include <stdarg.h>
#if defined(_WIN32)
#define WIN32_LEAN_AND_MEAN
@ -30,26 +40,44 @@
#define ANSI_COLOR_BLUE "\x1b[34m"
#define ANSI_COLOR_MAGENTA "\x1b[35m"
#define ANSI_COLOR_CYAN "\x1b[36m"
#define ANSI_COLOR_GRAY "\x1b[90m"
#define ANSI_COLOR_RESET "\x1b[0m"
#define ANSI_BOLD "\x1b[1m"
namespace console {
#if defined (_WIN32)
namespace {
// Use private-use unicode values to represent special keys that are not reported
// as characters (e.g. arrows on Windows). These values should never clash with
// real input and let the rest of the code handle navigation uniformly.
static constexpr char32_t KEY_ARROW_LEFT = 0xE000;
static constexpr char32_t KEY_ARROW_RIGHT = 0xE001;
static constexpr char32_t KEY_ARROW_UP = 0xE002;
static constexpr char32_t KEY_ARROW_DOWN = 0xE003;
static constexpr char32_t KEY_HOME = 0xE004;
static constexpr char32_t KEY_END = 0xE005;
static constexpr char32_t KEY_CTRL_ARROW_LEFT = 0xE006;
static constexpr char32_t KEY_CTRL_ARROW_RIGHT = 0xE007;
static constexpr char32_t KEY_DELETE = 0xE008;
}
//
// Console state
//
#endif
static bool advanced_display = false;
static bool simple_io = true;
static display_t current_display = reset;
static bool advanced_display = false;
static bool simple_io = true;
static display_type current_display = DISPLAY_TYPE_RESET;
static FILE* out = stdout;
static FILE* out = stdout;
#if defined (_WIN32)
static void* hConsole;
static void* hConsole;
#else
static FILE* tty = nullptr;
static termios initial_state;
static FILE* tty = nullptr;
static termios initial_state;
#endif
//
@ -120,7 +148,7 @@ namespace console {
void cleanup() {
// Reset console display
set_display(reset);
set_display(DISPLAY_TYPE_RESET);
#if !defined(_WIN32)
// Restore settings on POSIX systems
@ -140,20 +168,26 @@ namespace console {
//
// Keep track of current display and only emit ANSI code if it changes
void set_display(display_t display) {
void set_display(display_type display) {
if (advanced_display && current_display != display) {
fflush(stdout);
common_log_flush(common_log_main());
switch(display) {
case reset:
case DISPLAY_TYPE_RESET:
fprintf(out, ANSI_COLOR_RESET);
break;
case prompt:
case DISPLAY_TYPE_INFO:
fprintf(out, ANSI_COLOR_MAGENTA);
break;
case DISPLAY_TYPE_PROMPT:
fprintf(out, ANSI_COLOR_YELLOW);
break;
case user_input:
case DISPLAY_TYPE_REASONING:
fprintf(out, ANSI_COLOR_GRAY);
break;
case DISPLAY_TYPE_USER_INPUT:
fprintf(out, ANSI_BOLD ANSI_COLOR_GREEN);
break;
case error:
case DISPLAY_TYPE_ERROR:
fprintf(out, ANSI_BOLD ANSI_COLOR_RED);
}
current_display = display;
@ -176,7 +210,18 @@ namespace console {
if (record.EventType == KEY_EVENT && record.Event.KeyEvent.bKeyDown) {
wchar_t wc = record.Event.KeyEvent.uChar.UnicodeChar;
if (wc == 0) {
continue;
const DWORD ctrl_mask = LEFT_CTRL_PRESSED | RIGHT_CTRL_PRESSED;
const bool ctrl_pressed = (record.Event.KeyEvent.dwControlKeyState & ctrl_mask) != 0;
switch (record.Event.KeyEvent.wVirtualKeyCode) {
case VK_LEFT: return ctrl_pressed ? KEY_CTRL_ARROW_LEFT : KEY_ARROW_LEFT;
case VK_RIGHT: return ctrl_pressed ? KEY_CTRL_ARROW_RIGHT : KEY_ARROW_RIGHT;
case VK_UP: return KEY_ARROW_UP;
case VK_DOWN: return KEY_ARROW_DOWN;
case VK_HOME: return KEY_HOME;
case VK_END: return KEY_END;
case VK_DELETE: return KEY_DELETE;
default: continue;
}
}
if ((wc >= 0xD800) && (wc <= 0xDBFF)) { // Check if wc is a high surrogate
@ -315,6 +360,52 @@ namespace console {
#endif
}
static char32_t decode_utf8(const std::string & input, size_t pos, size_t & advance) {
unsigned char c = static_cast<unsigned char>(input[pos]);
if ((c & 0x80u) == 0u) {
advance = 1;
return c;
}
if ((c & 0xE0u) == 0xC0u && pos + 1 < input.size()) {
unsigned char c1 = static_cast<unsigned char>(input[pos + 1]);
if ((c1 & 0xC0u) != 0x80u) {
advance = 1;
return 0xFFFD;
}
advance = 2;
return ((c & 0x1Fu) << 6) | (static_cast<unsigned char>(input[pos + 1]) & 0x3Fu);
}
if ((c & 0xF0u) == 0xE0u && pos + 2 < input.size()) {
unsigned char c1 = static_cast<unsigned char>(input[pos + 1]);
unsigned char c2 = static_cast<unsigned char>(input[pos + 2]);
if ((c1 & 0xC0u) != 0x80u || (c2 & 0xC0u) != 0x80u) {
advance = 1;
return 0xFFFD;
}
advance = 3;
return ((c & 0x0Fu) << 12) |
((static_cast<unsigned char>(input[pos + 1]) & 0x3Fu) << 6) |
(static_cast<unsigned char>(input[pos + 2]) & 0x3Fu);
}
if ((c & 0xF8u) == 0xF0u && pos + 3 < input.size()) {
unsigned char c1 = static_cast<unsigned char>(input[pos + 1]);
unsigned char c2 = static_cast<unsigned char>(input[pos + 2]);
unsigned char c3 = static_cast<unsigned char>(input[pos + 3]);
if ((c1 & 0xC0u) != 0x80u || (c2 & 0xC0u) != 0x80u || (c3 & 0xC0u) != 0x80u) {
advance = 1;
return 0xFFFD;
}
advance = 4;
return ((c & 0x07u) << 18) |
((static_cast<unsigned char>(input[pos + 1]) & 0x3Fu) << 12) |
((static_cast<unsigned char>(input[pos + 2]) & 0x3Fu) << 6) |
(static_cast<unsigned char>(input[pos + 3]) & 0x3Fu);
}
advance = 1;
return 0xFFFD; // replacement character for invalid input
}
static void append_utf8(char32_t ch, std::string & out) {
if (ch <= 0x7F) {
out.push_back(static_cast<unsigned char>(ch));
@ -336,22 +427,319 @@ namespace console {
}
// Helper function to remove the last UTF-8 character from a string
static void pop_back_utf8_char(std::string & line) {
if (line.empty()) {
static size_t prev_utf8_char_pos(const std::string & line, size_t pos) {
if (pos == 0) return 0;
pos--;
while (pos > 0 && (line[pos] & 0xC0) == 0x80) {
pos--;
}
return pos;
}
static size_t next_utf8_char_pos(const std::string & line, size_t pos) {
if (pos >= line.length()) return line.length();
pos++;
while (pos < line.length() && (line[pos] & 0xC0) == 0x80) {
pos++;
}
return pos;
}
static void move_cursor(int delta);
static void move_word_left(size_t & char_pos, size_t & byte_pos, const std::vector<int> & widths, const std::string & line);
static void move_word_right(size_t & char_pos, size_t & byte_pos, const std::vector<int> & widths, const std::string & line);
static void move_to_line_start(size_t & char_pos, size_t & byte_pos, const std::vector<int> & widths);
static void move_to_line_end(size_t & char_pos, size_t & byte_pos, const std::vector<int> & widths, const std::string & line);
static void delete_at_cursor(std::string & line, std::vector<int> & widths, size_t & char_pos, size_t & byte_pos) {
if (char_pos >= widths.size()) {
return;
}
size_t pos = line.length() - 1;
size_t next_pos = next_utf8_char_pos(line, byte_pos);
int w = widths[char_pos];
size_t char_len = next_pos - byte_pos;
// Find the start of the last UTF-8 character (checking up to 4 bytes back)
for (size_t i = 0; i < 3 && pos > 0; ++i, --pos) {
if ((line[pos] & 0xC0) != 0x80) {
break; // Found the start of the character
}
line.erase(byte_pos, char_len);
widths.erase(widths.begin() + char_pos);
size_t p = byte_pos;
int tail_width = 0;
for (size_t i = char_pos; i < widths.size(); ++i) {
size_t following = next_utf8_char_pos(line, p);
put_codepoint(line.c_str() + p, following - p, widths[i]);
tail_width += widths[i];
p = following;
}
line.erase(pos);
for (int i = 0; i < w; ++i) {
fputc(' ', out);
}
move_cursor(-(tail_width + w));
}
static void clear_current_line(const std::vector<int> & widths) {
int total_width = 0;
for (int w : widths) {
total_width += (w > 0 ? w : 1);
}
if (total_width > 0) {
std::string spaces(total_width, ' ');
fwrite(spaces.c_str(), 1, total_width, out);
move_cursor(-total_width);
}
}
static void set_line_contents(std::string new_line, std::string & line, std::vector<int> & widths, size_t & char_pos,
size_t & byte_pos) {
move_to_line_start(char_pos, byte_pos, widths);
clear_current_line(widths);
line = std::move(new_line);
widths.clear();
byte_pos = 0;
char_pos = 0;
size_t idx = 0;
while (idx < line.size()) {
size_t advance = 0;
char32_t cp = decode_utf8(line, idx, advance);
int expected_width = estimateWidth(cp);
int real_width = put_codepoint(line.c_str() + idx, advance, expected_width);
if (real_width < 0) real_width = 0;
widths.push_back(real_width);
idx += advance;
++char_pos;
byte_pos = idx;
}
}
static void move_to_line_start(size_t & char_pos, size_t & byte_pos, const std::vector<int> & widths) {
int back_width = 0;
for (size_t i = 0; i < char_pos; ++i) {
back_width += widths[i];
}
move_cursor(-back_width);
char_pos = 0;
byte_pos = 0;
}
static void move_to_line_end(size_t & char_pos, size_t & byte_pos, const std::vector<int> & widths, const std::string & line) {
int forward_width = 0;
for (size_t i = char_pos; i < widths.size(); ++i) {
forward_width += widths[i];
}
move_cursor(forward_width);
char_pos = widths.size();
byte_pos = line.length();
}
static bool has_ctrl_modifier(const std::string & params) {
size_t start = 0;
while (start < params.size()) {
size_t end = params.find(';', start);
size_t len = (end == std::string::npos) ? params.size() - start : end - start;
if (len > 0) {
int value = 0;
for (size_t i = 0; i < len; ++i) {
char ch = params[start + i];
if (!std::isdigit(static_cast<unsigned char>(ch))) {
value = -1;
break;
}
value = value * 10 + (ch - '0');
}
if (value == 5) {
return true;
}
}
if (end == std::string::npos) {
break;
}
start = end + 1;
}
return false;
}
static bool is_space_codepoint(char32_t cp) {
return std::iswspace(static_cast<wint_t>(cp)) != 0;
}
static void move_word_left(size_t & char_pos, size_t & byte_pos, const std::vector<int> & widths, const std::string & line) {
if (char_pos == 0) {
return;
}
size_t new_char_pos = char_pos;
size_t new_byte_pos = byte_pos;
int move_width = 0;
while (new_char_pos > 0) {
size_t prev_byte = prev_utf8_char_pos(line, new_byte_pos);
size_t advance = 0;
char32_t cp = decode_utf8(line, prev_byte, advance);
if (!is_space_codepoint(cp)) {
break;
}
move_width += widths[new_char_pos - 1];
new_char_pos--;
new_byte_pos = prev_byte;
}
while (new_char_pos > 0) {
size_t prev_byte = prev_utf8_char_pos(line, new_byte_pos);
size_t advance = 0;
char32_t cp = decode_utf8(line, prev_byte, advance);
if (is_space_codepoint(cp)) {
break;
}
move_width += widths[new_char_pos - 1];
new_char_pos--;
new_byte_pos = prev_byte;
}
move_cursor(-move_width);
char_pos = new_char_pos;
byte_pos = new_byte_pos;
}
static void move_word_right(size_t & char_pos, size_t & byte_pos, const std::vector<int> & widths, const std::string & line) {
if (char_pos >= widths.size()) {
return;
}
size_t new_char_pos = char_pos;
size_t new_byte_pos = byte_pos;
int move_width = 0;
while (new_char_pos < widths.size()) {
size_t advance = 0;
char32_t cp = decode_utf8(line, new_byte_pos, advance);
if (!is_space_codepoint(cp)) {
break;
}
move_width += widths[new_char_pos];
new_char_pos++;
new_byte_pos += advance;
}
while (new_char_pos < widths.size()) {
size_t advance = 0;
char32_t cp = decode_utf8(line, new_byte_pos, advance);
if (is_space_codepoint(cp)) {
break;
}
move_width += widths[new_char_pos];
new_char_pos++;
new_byte_pos += advance;
}
while (new_char_pos < widths.size()) {
size_t advance = 0;
char32_t cp = decode_utf8(line, new_byte_pos, advance);
if (!is_space_codepoint(cp)) {
break;
}
move_width += widths[new_char_pos];
new_char_pos++;
new_byte_pos += advance;
}
move_cursor(move_width);
char_pos = new_char_pos;
byte_pos = new_byte_pos;
}
static void move_cursor(int delta) {
if (delta == 0) return;
#if defined(_WIN32)
if (hConsole != NULL) {
CONSOLE_SCREEN_BUFFER_INFO bufferInfo;
GetConsoleScreenBufferInfo(hConsole, &bufferInfo);
COORD newCursorPosition = bufferInfo.dwCursorPosition;
int width = bufferInfo.dwSize.X;
int newX = newCursorPosition.X + delta;
int newY = newCursorPosition.Y;
while (newX >= width) {
newX -= width;
newY++;
}
while (newX < 0) {
newX += width;
newY--;
}
newCursorPosition.X = newX;
newCursorPosition.Y = newY;
SetConsoleCursorPosition(hConsole, newCursorPosition);
}
#else
if (delta < 0) {
for (int i = 0; i < -delta; i++) fprintf(out, "\b");
} else {
for (int i = 0; i < delta; i++) fprintf(out, "\033[C");
}
#endif
}
struct history_t {
std::vector<std::string> entries;
size_t viewing_idx = SIZE_MAX;
std::string backup_line; // current line before viewing history
void add(const std::string & line) {
if (line.empty()) {
return;
}
// avoid duplicates with the last entry
if (entries.empty() || entries.back() != line) {
entries.push_back(line);
}
// also clear viewing state
end_viewing();
}
bool prev(std::string & cur_line) {
if (entries.empty()) {
return false;
}
if (viewing_idx == SIZE_MAX) {
return false;
}
if (viewing_idx > 0) {
viewing_idx--;
}
cur_line = entries[viewing_idx];
return true;
}
bool next(std::string & cur_line) {
if (entries.empty() || viewing_idx == SIZE_MAX) {
return false;
}
viewing_idx++;
if (viewing_idx >= entries.size()) {
cur_line = backup_line;
end_viewing();
} else {
cur_line = entries[viewing_idx];
}
return true;
}
void begin_viewing(const std::string & line) {
backup_line = line;
viewing_idx = entries.size();
}
void end_viewing() {
viewing_idx = SIZE_MAX;
backup_line.clear();
}
bool is_viewing() const {
return viewing_idx != SIZE_MAX;
}
} history;
static bool readline_advanced(std::string & line, bool multiline_input) {
if (out != stdout) {
fflush(stdout);
@ -362,8 +750,33 @@ namespace console {
bool is_special_char = false;
bool end_of_stream = false;
size_t byte_pos = 0; // current byte index
size_t char_pos = 0; // current character index (one char can be multiple bytes)
char32_t input_char;
while (true) {
assert(char_pos <= byte_pos);
assert(char_pos <= widths.size());
auto history_prev = [&]() {
if (!history.is_viewing()) {
history.begin_viewing(line);
}
std::string new_line;
if (!history.prev(new_line)) {
return;
}
set_line_contents(new_line, line, widths, char_pos, byte_pos);
};
auto history_next = [&]() {
if (history.is_viewing()) {
std::string new_line;
if (!history.next(new_line)) {
return;
}
set_line_contents(new_line, line, widths, char_pos, byte_pos);
}
};
fflush(out); // Ensure all output is displayed before waiting for input
input_char = getchar32();
@ -371,20 +784,83 @@ namespace console {
break;
}
if (input_char == (char32_t) WEOF || input_char == 0x04 /* Ctrl+D*/) {
if (input_char == (char32_t) WEOF || input_char == 0x04 /* Ctrl+D */) {
end_of_stream = true;
break;
}
if (is_special_char) {
set_display(user_input);
replace_last(line.back());
is_special_char = false;
}
if (input_char == '\033') { // Escape sequence
char32_t code = getchar32();
if (code == '[' || code == 0x1B) {
if (code == '[') {
std::string params;
while (true) {
code = getchar32();
if ((code >= 'A' && code <= 'Z') || (code >= 'a' && code <= 'z') || code == '~' || code == (char32_t) WEOF) {
break;
}
params.push_back(static_cast<char>(code));
}
const bool ctrl_modifier = has_ctrl_modifier(params);
if (code == 'D') { // left
if (ctrl_modifier) {
move_word_left(char_pos, byte_pos, widths, line);
} else if (char_pos > 0) {
int w = widths[char_pos - 1];
move_cursor(-w);
char_pos--;
byte_pos = prev_utf8_char_pos(line, byte_pos);
}
} else if (code == 'C') { // right
if (ctrl_modifier) {
move_word_right(char_pos, byte_pos, widths, line);
} else if (char_pos < widths.size()) {
int w = widths[char_pos];
move_cursor(w);
char_pos++;
byte_pos = next_utf8_char_pos(line, byte_pos);
}
} else if (code == 'H') { // home
move_to_line_start(char_pos, byte_pos, widths);
} else if (code == 'F') { // end
move_to_line_end(char_pos, byte_pos, widths, line);
} else if (code == 'A' || code == 'B') {
// up/down
if (code == 'A') {
history_prev();
is_special_char = false;
} else if (code == 'B') {
history_next();
is_special_char = false;
}
} else if ((code == '~' || (code >= 'A' && code <= 'Z') || (code >= 'a' && code <= 'z')) && !params.empty()) {
std::string digits;
for (char ch : params) {
if (ch == ';') {
break;
}
if (std::isdigit(static_cast<unsigned char>(ch))) {
digits.push_back(ch);
}
}
if (code == '~') {
if (digits == "1" || digits == "7") { // home
move_to_line_start(char_pos, byte_pos, widths);
} else if (digits == "4" || digits == "8") { // end
move_to_line_end(char_pos, byte_pos, widths, line);
} else if (digits == "3") { // delete
delete_at_cursor(line, widths, char_pos, byte_pos);
}
}
}
} else if (code == 0x1B) {
// Discard the rest of the escape sequence
while ((code = getchar32()) != (char32_t) WEOF) {
if ((code >= 'A' && code <= 'Z') || (code >= 'a' && code <= 'z') || code == '~') {
@ -392,32 +868,110 @@ namespace console {
}
}
}
#if defined(_WIN32)
} else if (input_char == KEY_ARROW_LEFT) {
if (char_pos > 0) {
int w = widths[char_pos - 1];
move_cursor(-w);
char_pos--;
byte_pos = prev_utf8_char_pos(line, byte_pos);
}
} else if (input_char == KEY_ARROW_RIGHT) {
if (char_pos < widths.size()) {
int w = widths[char_pos];
move_cursor(w);
char_pos++;
byte_pos = next_utf8_char_pos(line, byte_pos);
}
} else if (input_char == KEY_CTRL_ARROW_LEFT) {
move_word_left(char_pos, byte_pos, widths, line);
} else if (input_char == KEY_CTRL_ARROW_RIGHT) {
move_word_right(char_pos, byte_pos, widths, line);
} else if (input_char == KEY_HOME) {
move_to_line_start(char_pos, byte_pos, widths);
} else if (input_char == KEY_END) {
move_to_line_end(char_pos, byte_pos, widths, line);
} else if (input_char == KEY_DELETE) {
delete_at_cursor(line, widths, char_pos, byte_pos);
} else if (input_char == KEY_ARROW_UP || input_char == KEY_ARROW_DOWN) {
if (input_char == KEY_ARROW_UP) {
history_prev();
is_special_char = false;
} else if (input_char == KEY_ARROW_DOWN) {
history_next();
is_special_char = false;
}
#endif
} else if (input_char == 0x08 || input_char == 0x7F) { // Backspace
if (!widths.empty()) {
int count;
do {
count = widths.back();
widths.pop_back();
// Move cursor back, print space, and move cursor back again
for (int i = 0; i < count; i++) {
replace_last(' ');
pop_cursor();
}
pop_back_utf8_char(line);
} while (count == 0 && !widths.empty());
if (char_pos > 0) {
int w = widths[char_pos - 1];
move_cursor(-w);
char_pos--;
size_t prev_pos = prev_utf8_char_pos(line, byte_pos);
size_t char_len = byte_pos - prev_pos;
byte_pos = prev_pos;
// remove the character
line.erase(byte_pos, char_len);
widths.erase(widths.begin() + char_pos);
// redraw tail
size_t p = byte_pos;
int tail_width = 0;
for (size_t i = char_pos; i < widths.size(); ++i) {
size_t next_p = next_utf8_char_pos(line, p);
put_codepoint(line.c_str() + p, next_p - p, widths[i]);
tail_width += widths[i];
p = next_p;
}
// clear display
for (int i = 0; i < w; ++i) {
fputc(' ', out);
}
move_cursor(-(tail_width + w));
}
} else {
int offset = line.length();
append_utf8(input_char, line);
int width = put_codepoint(line.c_str() + offset, line.length() - offset, estimateWidth(input_char));
if (width < 0) {
width = 0;
// insert character
std::string new_char_str;
append_utf8(input_char, new_char_str);
int w = estimateWidth(input_char);
if (char_pos == widths.size()) {
// insert at the end
line += new_char_str;
int real_w = put_codepoint(new_char_str.c_str(), new_char_str.length(), w);
if (real_w < 0) real_w = 0;
widths.push_back(real_w);
byte_pos += new_char_str.length();
char_pos++;
} else {
// insert in middle
line.insert(byte_pos, new_char_str);
int real_w = put_codepoint(new_char_str.c_str(), new_char_str.length(), w);
if (real_w < 0) real_w = 0;
widths.insert(widths.begin() + char_pos, real_w);
// print the tail
size_t p = byte_pos + new_char_str.length();
int tail_width = 0;
for (size_t i = char_pos + 1; i < widths.size(); ++i) {
size_t next_p = next_utf8_char_pos(line, p);
put_codepoint(line.c_str() + p, next_p - p, widths[i]);
tail_width += widths[i];
p = next_p;
}
move_cursor(-tail_width);
byte_pos += new_char_str.length();
char_pos++;
}
widths.push_back(width);
}
if (!line.empty() && (line.back() == '\\' || line.back() == '/')) {
set_display(prompt);
replace_last(line.back());
is_special_char = true;
}
@ -451,6 +1005,15 @@ namespace console {
}
}
if (!end_of_stream && !line.empty()) {
// remove the trailing newline for history storage
if (!line.empty() && line.back() == '\n') {
line.pop_back();
}
// TODO: maybe support multiline history entries?
history.add(line);
}
fflush(out);
return has_more;
}
@ -493,12 +1056,82 @@ namespace console {
}
bool readline(std::string & line, bool multiline_input) {
set_display(user_input);
if (simple_io) {
return readline_simple(line, multiline_input);
}
return readline_advanced(line, multiline_input);
}
namespace spinner {
static const char LOADING_CHARS[] = {'|', '/', '-', '\\'};
static std::condition_variable cv_stop;
static std::thread th;
static size_t frame = 0; // only modified by one thread
static bool running = false;
static std::mutex mtx;
static auto wait_time = std::chrono::milliseconds(100);
static void draw_next_frame() {
// don't need lock because only one thread modifies running
frame = (frame + 1) % sizeof(LOADING_CHARS);
replace_last(LOADING_CHARS[frame]);
fflush(out);
}
void start() {
std::unique_lock<std::mutex> lock(mtx);
if (simple_io || running) {
return;
}
common_log_flush(common_log_main());
fprintf(out, "%c", LOADING_CHARS[0]);
fflush(out);
frame = 1;
running = true;
th = std::thread([]() {
std::unique_lock<std::mutex> lock(mtx);
while (true) {
if (cv_stop.wait_for(lock, wait_time, []{ return !running; })) {
break;
}
draw_next_frame();
}
});
}
void stop() {
{
std::unique_lock<std::mutex> lock(mtx);
if (simple_io || !running) {
return;
}
running = false;
cv_stop.notify_all();
}
if (th.joinable()) {
th.join();
}
replace_last(' ');
pop_cursor();
fflush(out);
}
}
void log(const char * fmt, ...) {
va_list args;
va_start(args, fmt);
vfprintf(out, fmt, args);
va_end(args);
}
void error(const char * fmt, ...) {
va_list args;
va_start(args, fmt);
display_type cur = current_display;
set_display(DISPLAY_TYPE_ERROR);
vfprintf(out, fmt, args);
set_display(cur); // restore previous color
va_end(args);
}
void flush() {
fflush(out);
}
}

38
common/console.h
View File

@ -2,18 +2,40 @@
#pragma once
#include "common.h"
#include <string>
namespace console {
enum display_t {
reset = 0,
prompt,
user_input,
error
};
enum display_type {
DISPLAY_TYPE_RESET = 0,
DISPLAY_TYPE_INFO,
DISPLAY_TYPE_PROMPT,
DISPLAY_TYPE_REASONING,
DISPLAY_TYPE_USER_INPUT,
DISPLAY_TYPE_ERROR
};
namespace console {
void init(bool use_simple_io, bool use_advanced_display);
void cleanup();
void set_display(display_t display);
void set_display(display_type display);
bool readline(std::string & line, bool multiline_input);
namespace spinner {
void start();
void stop();
}
// note: the logging API below output directly to stdout
// it can negatively impact performance if used on inference thread
// only use in in a dedicated CLI thread
// for logging in inference thread, use log.h instead
LLAMA_COMMON_ATTRIBUTE_FORMAT(1, 2)
void log(const char * fmt, ...);
LLAMA_COMMON_ATTRIBUTE_FORMAT(1, 2)
void error(const char * fmt, ...);
void flush();
}

118
common/download.cpp
View File

@ -12,6 +12,8 @@
#include <filesystem>
#include <fstream>
#include <future>
#include <map>
#include <mutex>
#include <regex>
#include <string>
#include <thread>
@ -24,6 +26,7 @@
#include "http.h"
#endif
#ifndef __EMSCRIPTEN__
#ifdef __linux__
#include <linux/limits.h>
#elif defined(_WIN32)
@ -35,6 +38,8 @@
#else
#include <sys/syslimits.h>
#endif
#endif
#define LLAMA_MAX_URL_LENGTH 2084 // Maximum URL Length in Chrome: 2083
// isatty
@ -430,7 +435,7 @@ std::pair<long, std::vector<char>> common_remote_get_content(const std::string &
curl_easy_setopt(curl.get(), CURLOPT_URL, url.c_str());
curl_easy_setopt(curl.get(), CURLOPT_NOPROGRESS, 1L);
curl_easy_setopt(curl.get(), CURLOPT_FOLLOWLOCATION, 1L);
curl_easy_setopt(curl.get(), CURLOPT_VERBOSE, 1L);
curl_easy_setopt(curl.get(), CURLOPT_VERBOSE, 0L);
typedef size_t(*CURLOPT_WRITEFUNCTION_PTR)(void * ptr, size_t size, size_t nmemb, void * data);
auto write_callback = [](void * ptr, size_t size, size_t nmemb, void * data) -> size_t {
auto data_vec = static_cast<std::vector<char> *>(data);
@ -469,36 +474,79 @@ std::pair<long, std::vector<char>> common_remote_get_content(const std::string &
#elif defined(LLAMA_USE_HTTPLIB)
static bool is_output_a_tty() {
class ProgressBar {
static inline std::mutex mutex;
static inline std::map<const ProgressBar *, int> lines;
static inline int max_line = 0;
static void cleanup(const ProgressBar * line) {
lines.erase(line);
if (lines.empty()) {
max_line = 0;
}
}
static bool is_output_a_tty() {
#if defined(_WIN32)
return _isatty(_fileno(stdout));
return _isatty(_fileno(stdout));
#else
return isatty(1);
return isatty(1);
#endif
}
static void print_progress(size_t current, size_t total) {
if (!is_output_a_tty()) {
return;
}
if (!total) {
return;
public:
ProgressBar() = default;
~ProgressBar() {
std::lock_guard<std::mutex> lock(mutex);
cleanup(this);
}
size_t width = 50;
size_t pct = (100 * current) / total;
size_t pos = (width * current) / total;
void update(size_t current, size_t total) {
if (!is_output_a_tty()) {
return;
}
std::cout << "["
<< std::string(pos, '=')
<< (pos < width ? ">" : "")
<< std::string(width - pos, ' ')
<< "] " << std::setw(3) << pct << "% ("
<< current / (1024 * 1024) << " MB / "
<< total / (1024 * 1024) << " MB)\r";
std::cout.flush();
}
if (!total) {
return;
}
std::lock_guard<std::mutex> lock(mutex);
if (lines.find(this) == lines.end()) {
lines[this] = max_line++;
std::cout << "\n";
}
int lines_up = max_line - lines[this];
size_t width = 50;
size_t pct = (100 * current) / total;
size_t pos = (width * current) / total;
std::cout << "\033[s";
if (lines_up > 0) {
std::cout << "\033[" << lines_up << "A";
}
std::cout << "\033[2K\r["
<< std::string(pos, '=')
<< (pos < width ? ">" : "")
<< std::string(width - pos, ' ')
<< "] " << std::setw(3) << pct << "% ("
<< current / (1024 * 1024) << " MB / "
<< total / (1024 * 1024) << " MB) "
<< "\033[u";
std::cout.flush();
if (current == total) {
cleanup(this);
}
}
ProgressBar(const ProgressBar &) = delete;
ProgressBar & operator=(const ProgressBar &) = delete;
};
static bool common_pull_file(httplib::Client & cli,
const std::string & resolve_path,
@ -517,16 +565,19 @@ static bool common_pull_file(httplib::Client & cli,
headers.emplace("Range", "bytes=" + std::to_string(existing_size) + "-");
}
std::atomic<size_t> downloaded{existing_size};
const char * func = __func__; // avoid __func__ inside a lambda
size_t downloaded = existing_size;
size_t progress_step = 0;
ProgressBar bar;
auto res = cli.Get(resolve_path, headers,
[&](const httplib::Response &response) {
if (existing_size > 0 && response.status != 206) {
LOG_WRN("%s: server did not respond with 206 Partial Content for a resume request. Status: %d\n", __func__, response.status);
LOG_WRN("%s: server did not respond with 206 Partial Content for a resume request. Status: %d\n", func, response.status);
return false;
}
if (existing_size == 0 && response.status != 200) {
LOG_WRN("%s: download received non-successful status code: %d\n", __func__, response.status);
LOG_WRN("%s: download received non-successful status code: %d\n", func, response.status);
return false;
}
if (total_size == 0 && response.has_header("Content-Length")) {
@ -534,7 +585,7 @@ static bool common_pull_file(httplib::Client & cli,
size_t content_length = std::stoull(response.get_header_value("Content-Length"));
total_size = existing_size + content_length;
} catch (const std::exception &e) {
LOG_WRN("%s: invalid Content-Length header: %s\n", __func__, e.what());
LOG_WRN("%s: invalid Content-Length header: %s\n", func, e.what());
}
}
return true;
@ -542,18 +593,21 @@ static bool common_pull_file(httplib::Client & cli,
[&](const char *data, size_t len) {
ofs.write(data, len);
if (!ofs) {
LOG_ERR("%s: error writing to file: %s\n", __func__, path_tmp.c_str());
LOG_ERR("%s: error writing to file: %s\n", func, path_tmp.c_str());
return false;
}
downloaded += len;
print_progress(downloaded, total_size);
progress_step += len;
if (progress_step >= total_size / 1000 || downloaded == total_size) {
bar.update(downloaded, total_size);
progress_step = 0;
}
return true;
},
nullptr
);
std::cout << "\n";
if (!res) {
LOG_ERR("%s: error during download. Status: %d\n", __func__, res ? res->status : -1);
return false;
@ -1047,7 +1101,7 @@ std::string common_docker_resolve_model(const std::string &) {
std::vector<common_cached_model_info> common_list_cached_models() {
std::vector<common_cached_model_info> models;
const std::string cache_dir = fs_get_cache_directory();
const std::vector<common_file_info> files = fs_list_files(cache_dir);
const std::vector<common_file_info> files = fs_list(cache_dir, false);
for (const auto & file : files) {
if (string_starts_with(file.name, "manifest=") && string_ends_with(file.name, ".json")) {
common_cached_model_info model_info;

4
common/download.h
View File

@ -14,8 +14,10 @@ struct common_cached_model_info {
std::string model;
std::string tag;
size_t size = 0; // GGUF size in bytes
// return string representation like "user/model:tag"
// if tag is "latest", it will be omitted
std::string to_string() const {
return user + "/" + model + ":" + tag;
return user + "/" + model + (tag == "latest" ? "" : ":" + tag);
}
};

6
common/json-schema-to-grammar.cpp
View File

@ -268,10 +268,10 @@ static bool is_reserved_name(const std::string & name) {
}
std::regex INVALID_RULE_CHARS_RE("[^a-zA-Z0-9-]+");
std::regex GRAMMAR_LITERAL_ESCAPE_RE("[\r\n\"]");
std::regex GRAMMAR_LITERAL_ESCAPE_RE("[\r\n\"\\\\]");
std::regex GRAMMAR_RANGE_LITERAL_ESCAPE_RE("[\r\n\"\\]\\-\\\\]");
std::unordered_map<char, std::string> GRAMMAR_LITERAL_ESCAPES = {
{'\r', "\\r"}, {'\n', "\\n"}, {'"', "\\\""}, {'-', "\\-"}, {']', "\\]"}
{'\r', "\\r"}, {'\n', "\\n"}, {'"', "\\\""}, {'-', "\\-"}, {']', "\\]"}, {'\\', "\\\\"}
};
std::unordered_set<char> NON_LITERAL_SET = {'|', '.', '(', ')', '[', ']', '{', '}', '*', '+', '?'};
@ -974,7 +974,7 @@ public:
void check_errors() {
if (!_errors.empty()) {
throw std::runtime_error("JSON schema conversion failed:\n" + string_join(_errors, "\n"));
throw std::invalid_argument("JSON schema conversion failed:\n" + string_join(_errors, "\n"));
}
if (!_warnings.empty()) {
fprintf(stderr, "WARNING: JSON schema conversion was incomplete: %s\n", string_join(_warnings, "; ").c_str());

50
common/log.cpp
View File

@ -1,3 +1,4 @@
#include "common.h"
#include "log.h"
#include <chrono>
@ -26,30 +27,6 @@ void common_log_set_verbosity_thold(int verbosity) {
common_log_verbosity_thold = verbosity;
}
// Auto-detect if colors should be enabled based on terminal and environment
static bool common_log_should_use_colors_auto() {
// Check NO_COLOR environment variable (https://no-color.org/)
if (const char * no_color = std::getenv("NO_COLOR")) {
if (no_color[0] != '\0') {
return false;
}
}
// Check TERM environment variable
if (const char * term = std::getenv("TERM")) {
if (std::strcmp(term, "dumb") == 0) {
return false;
}
}
// Check if stdout and stderr are connected to a terminal
// We check both because log messages can go to either
bool stdout_is_tty = isatty(fileno(stdout));
bool stderr_is_tty = isatty(fileno(stderr));
return stdout_is_tty || stderr_is_tty;
}
static int64_t t_us() {
return std::chrono::duration_cast<std::chrono::microseconds>(std::chrono::system_clock::now().time_since_epoch()).count();
}
@ -391,7 +368,7 @@ struct common_log * common_log_main() {
static std::once_flag init_flag;
std::call_once(init_flag, [&]() {
// Set default to auto-detect colors
log.set_colors(common_log_should_use_colors_auto());
log.set_colors(tty_can_use_colors());
});
return &log;
@ -422,7 +399,7 @@ void common_log_set_file(struct common_log * log, const char * file) {
void common_log_set_colors(struct common_log * log, log_colors colors) {
if (colors == LOG_COLORS_AUTO) {
log->set_colors(common_log_should_use_colors_auto());
log->set_colors(tty_can_use_colors());
return;
}
@ -443,8 +420,27 @@ void common_log_set_timestamps(struct common_log * log, bool timestamps) {
log->set_timestamps(timestamps);
}
void common_log_flush(struct common_log * log) {
log->pause();
log->resume();
}
static int common_get_verbosity(enum ggml_log_level level) {
switch (level) {
case GGML_LOG_LEVEL_DEBUG: return LOG_LEVEL_DEBUG;
case GGML_LOG_LEVEL_INFO: return LOG_LEVEL_INFO;
case GGML_LOG_LEVEL_WARN: return LOG_LEVEL_WARN;
case GGML_LOG_LEVEL_ERROR: return LOG_LEVEL_ERROR;
case GGML_LOG_LEVEL_CONT: return LOG_LEVEL_INFO; // same as INFO
case GGML_LOG_LEVEL_NONE:
default:
return LOG_LEVEL_OUTPUT;
}
}
void common_log_default_callback(enum ggml_log_level level, const char * text, void * /*user_data*/) {
if (LOG_DEFAULT_LLAMA <= common_log_verbosity_thold) {
auto verbosity = common_get_verbosity(level);
if (verbosity <= common_log_verbosity_thold) {
common_log_add(common_log_main(), level, "%s", text);
}
}

32
common/log.h
View File

@ -21,8 +21,14 @@
# define LOG_ATTRIBUTE_FORMAT(...) __attribute__((format(printf, __VA_ARGS__)))
#endif
#define LOG_DEFAULT_DEBUG 1
#define LOG_DEFAULT_LLAMA 0
#define LOG_LEVEL_DEBUG 4
#define LOG_LEVEL_INFO 3
#define LOG_LEVEL_WARN 2
#define LOG_LEVEL_ERROR 1
#define LOG_LEVEL_OUTPUT 0 // output data from tools
#define LOG_DEFAULT_DEBUG LOG_LEVEL_DEBUG
#define LOG_DEFAULT_LLAMA LOG_LEVEL_INFO
enum log_colors {
LOG_COLORS_AUTO = -1,
@ -67,16 +73,18 @@ void common_log_add(struct common_log * log, enum ggml_log_level level, const ch
// 0.00.090.578 I llm_load_tensors: offloading 32 repeating layers to GPU
// 0.00.090.579 I llm_load_tensors: offloading non-repeating layers to GPU
//
// I - info (stdout, V = 0)
// W - warning (stderr, V = 0)
// E - error (stderr, V = 0)
// D - debug (stderr, V = LOG_DEFAULT_DEBUG)
// I - info (stdout, V = LOG_DEFAULT_INFO)
// W - warning (stderr, V = LOG_DEFAULT_WARN)
// E - error (stderr, V = LOG_DEFAULT_ERROR)
// O - output (stdout, V = LOG_DEFAULT_OUTPUT)
//
void common_log_set_file (struct common_log * log, const char * file); // not thread-safe
void common_log_set_colors (struct common_log * log, log_colors colors); // not thread-safe
void common_log_set_prefix (struct common_log * log, bool prefix); // whether to output prefix to each log
void common_log_set_timestamps(struct common_log * log, bool timestamps); // whether to output timestamps in the prefix
void common_log_flush (struct common_log * log); // flush all pending log messages
// helper macros for logging
// use these to avoid computing log arguments if the verbosity of the log is higher than the threshold
@ -95,14 +103,14 @@ void common_log_set_timestamps(struct common_log * log, bool timestamps); // w
} \
} while (0)
#define LOG(...) LOG_TMPL(GGML_LOG_LEVEL_NONE, 0, __VA_ARGS__)
#define LOGV(verbosity, ...) LOG_TMPL(GGML_LOG_LEVEL_NONE, verbosity, __VA_ARGS__)
#define LOG(...) LOG_TMPL(GGML_LOG_LEVEL_NONE, LOG_LEVEL_OUTPUT, __VA_ARGS__)
#define LOGV(verbosity, ...) LOG_TMPL(GGML_LOG_LEVEL_NONE, verbosity, __VA_ARGS__)
#define LOG_INF(...) LOG_TMPL(GGML_LOG_LEVEL_INFO, 0, __VA_ARGS__)
#define LOG_WRN(...) LOG_TMPL(GGML_LOG_LEVEL_WARN, 0, __VA_ARGS__)
#define LOG_ERR(...) LOG_TMPL(GGML_LOG_LEVEL_ERROR, 0, __VA_ARGS__)
#define LOG_DBG(...) LOG_TMPL(GGML_LOG_LEVEL_DEBUG, LOG_DEFAULT_DEBUG, __VA_ARGS__)
#define LOG_CNT(...) LOG_TMPL(GGML_LOG_LEVEL_CONT, 0, __VA_ARGS__)
#define LOG_DBG(...) LOG_TMPL(GGML_LOG_LEVEL_DEBUG, LOG_LEVEL_DEBUG, __VA_ARGS__)
#define LOG_INF(...) LOG_TMPL(GGML_LOG_LEVEL_INFO, LOG_LEVEL_INFO, __VA_ARGS__)
#define LOG_WRN(...) LOG_TMPL(GGML_LOG_LEVEL_WARN, LOG_LEVEL_WARN, __VA_ARGS__)
#define LOG_ERR(...) LOG_TMPL(GGML_LOG_LEVEL_ERROR, LOG_LEVEL_ERROR, __VA_ARGS__)
#define LOG_CNT(...) LOG_TMPL(GGML_LOG_LEVEL_CONT, LOG_LEVEL_INFO, __VA_ARGS__) // same as INFO
#define LOG_INFV(verbosity, ...) LOG_TMPL(GGML_LOG_LEVEL_INFO, verbosity, __VA_ARGS__)
#define LOG_WRNV(verbosity, ...) LOG_TMPL(GGML_LOG_LEVEL_WARN, verbosity, __VA_ARGS__)

1712
common/peg-parser.cpp Normal file
View File

File diff suppressed because it is too large Load Diff

459
common/peg-parser.h Normal file
View File

@ -0,0 +1,459 @@
#pragma once
#include <nlohmann/json_fwd.hpp>
#include <memory>
#include <unordered_map>
#include <string>
#include <string_view>
#include <functional>
#include <vector>
#include <variant>
struct common_grammar_builder;
class common_peg_parser_builder;
using common_peg_parser_id = size_t;
constexpr common_peg_parser_id COMMON_PEG_INVALID_PARSER_ID = static_cast<common_peg_parser_id>(-1);
using common_peg_ast_id = size_t;
constexpr common_peg_ast_id COMMON_PEG_INVALID_AST_ID = static_cast<common_peg_ast_id>(-1);
// Lightweight wrapper around common_peg_parser_id for convenience
class common_peg_parser {
common_peg_parser_id id_;
common_peg_parser_builder & builder_;
public:
common_peg_parser(const common_peg_parser & other) : id_(other.id_), builder_(other.builder_) {}
common_peg_parser(common_peg_parser_id id, common_peg_parser_builder & builder) : id_(id), builder_(builder) {}
common_peg_parser & operator=(const common_peg_parser & other);
common_peg_parser & operator+=(const common_peg_parser & other);
common_peg_parser & operator|=(const common_peg_parser & other);
operator common_peg_parser_id() const { return id_; }
common_peg_parser_id id() const { return id_; }
common_peg_parser_builder & builder() const { return builder_; }
// Creates a sequence
common_peg_parser operator+(const common_peg_parser & other) const;
// Creates a sequence separated by spaces.
common_peg_parser operator<<(const common_peg_parser & other) const;
// Creates a choice
common_peg_parser operator|(const common_peg_parser & other) const;
common_peg_parser operator+(const char * str) const;
common_peg_parser operator+(const std::string & str) const;
common_peg_parser operator<<(const char * str) const;
common_peg_parser operator<<(const std::string & str) const;
common_peg_parser operator|(const char * str) const;
common_peg_parser operator|(const std::string & str) const;
};
common_peg_parser operator+(const char * str, const common_peg_parser & p);
common_peg_parser operator+(const std::string & str, const common_peg_parser & p);
common_peg_parser operator<<(const char * str, const common_peg_parser & p);
common_peg_parser operator<<(const std::string & str, const common_peg_parser & p);
common_peg_parser operator|(const char * str, const common_peg_parser & p);
common_peg_parser operator|(const std::string & str, const common_peg_parser & p);
enum common_peg_parse_result_type {
COMMON_PEG_PARSE_RESULT_FAIL = 0,
COMMON_PEG_PARSE_RESULT_SUCCESS = 1,
COMMON_PEG_PARSE_RESULT_NEED_MORE_INPUT = 2,
};
const char * common_peg_parse_result_type_name(common_peg_parse_result_type type);
struct common_peg_ast_node {
common_peg_ast_id id;
std::string rule;
std::string tag;
size_t start;
size_t end;
std::string_view text;
std::vector<common_peg_ast_id> children;
bool is_partial = false;
};
struct common_peg_parse_result;
using common_peg_ast_visitor = std::function<void(const common_peg_ast_node & node)>;
class common_peg_ast_arena {
std::vector<common_peg_ast_node> nodes_;
public:
common_peg_ast_id add_node(
const std::string & rule,
const std::string & tag,
size_t start,
size_t end,
std::string_view text,
std::vector<common_peg_ast_id> children,
bool is_partial = false
) {
common_peg_ast_id id = nodes_.size();
nodes_.push_back({id, rule, tag, start, end, text, std::move(children), is_partial});
return id;
}
const common_peg_ast_node & get(common_peg_ast_id id) const { return nodes_.at(id); }
size_t size() const { return nodes_.size(); }
void clear() { nodes_.clear(); }
void visit(common_peg_ast_id id, const common_peg_ast_visitor & visitor) const;
void visit(const common_peg_parse_result & result, const common_peg_ast_visitor & visitor) const;
};
struct common_peg_parse_result {
common_peg_parse_result_type type = COMMON_PEG_PARSE_RESULT_FAIL;
size_t start = 0;
size_t end = 0;
std::vector<common_peg_ast_id> nodes;
common_peg_parse_result() = default;
common_peg_parse_result(common_peg_parse_result_type type, size_t start)
: type(type), start(start), end(start) {}
common_peg_parse_result(common_peg_parse_result_type type, size_t start, size_t end)
: type(type), start(start), end(end) {}
common_peg_parse_result(common_peg_parse_result_type type, size_t start, size_t end, std::vector<common_peg_ast_id> nodes)
: type(type), start(start), end(end), nodes(std::move(nodes)) {}
bool fail() const { return type == COMMON_PEG_PARSE_RESULT_FAIL; }
bool need_more_input() const { return type == COMMON_PEG_PARSE_RESULT_NEED_MORE_INPUT; }
bool success() const { return type == COMMON_PEG_PARSE_RESULT_SUCCESS; }
};
struct common_peg_parse_context {
std::string input;
bool is_partial;
common_peg_ast_arena ast;
int parse_depth;
common_peg_parse_context()
: is_partial(false), parse_depth(0) {}
common_peg_parse_context(const std::string & input)
: input(input), is_partial(false), parse_depth(0) {}
common_peg_parse_context(const std::string & input, bool is_partial)
: input(input), is_partial(is_partial), parse_depth(0) {}
};
class common_peg_arena;
// Parser variants
struct common_peg_epsilon_parser {};
struct common_peg_start_parser {};
struct common_peg_end_parser {};
struct common_peg_literal_parser {
std::string literal;
};
struct common_peg_sequence_parser {
std::vector<common_peg_parser_id> children;
};
struct common_peg_choice_parser {
std::vector<common_peg_parser_id> children;
};
struct common_peg_repetition_parser {
common_peg_parser_id child;
int min_count;
int max_count; // -1 for unbounded
};
struct common_peg_and_parser {
common_peg_parser_id child;
};
struct common_peg_not_parser {
common_peg_parser_id child;
};
struct common_peg_any_parser {};
struct common_peg_space_parser {};
struct common_peg_chars_parser {
struct char_range {
uint32_t start;
uint32_t end;
bool contains(uint32_t codepoint) const { return codepoint >= start && codepoint <= end; }
};
std::string pattern;
std::vector<char_range> ranges;
bool negated;
int min_count;
int max_count; // -1 for unbounded
};
struct common_peg_json_string_parser {};
struct common_peg_until_parser {
std::vector<std::string> delimiters;
};
struct common_peg_schema_parser {
common_peg_parser_id child;
std::string name;
std::shared_ptr<nlohmann::ordered_json> schema;
// Indicates if the GBNF should accept a raw string that matches the schema.
bool raw;
};
struct common_peg_rule_parser {
std::string name;
common_peg_parser_id child;
bool trigger;
};
struct common_peg_ref_parser {
std::string name;
};
struct common_peg_atomic_parser {
common_peg_parser_id child;
};
struct common_peg_tag_parser {
common_peg_parser_id child;
std::string tag;
};
// Variant holding all parser types
using common_peg_parser_variant = std::variant<
common_peg_epsilon_parser,
common_peg_start_parser,
common_peg_end_parser,
common_peg_literal_parser,
common_peg_sequence_parser,
common_peg_choice_parser,
common_peg_repetition_parser,
common_peg_and_parser,
common_peg_not_parser,
common_peg_any_parser,
common_peg_space_parser,
common_peg_chars_parser,
common_peg_json_string_parser,
common_peg_until_parser,
common_peg_schema_parser,
common_peg_rule_parser,
common_peg_ref_parser,
common_peg_atomic_parser,
common_peg_tag_parser
>;
class common_peg_arena {
std::vector<common_peg_parser_variant> parsers_;
std::unordered_map<std::string, common_peg_parser_id> rules_;
common_peg_parser_id root_ = COMMON_PEG_INVALID_PARSER_ID;
public:
const common_peg_parser_variant & get(common_peg_parser_id id) const { return parsers_.at(id); }
common_peg_parser_variant & get(common_peg_parser_id id) { return parsers_.at(id); }
size_t size() const { return parsers_.size(); }
bool empty() const { return parsers_.empty(); }
common_peg_parser_id get_rule(const std::string & name) const;
bool has_rule(const std::string & name) const { return rules_.find(name) != rules_.end(); }
common_peg_parser_id root() const { return root_; }
void set_root(common_peg_parser_id id) { root_ = id; }
common_peg_parse_result parse(common_peg_parse_context & ctx, size_t start = 0) const;
common_peg_parse_result parse(common_peg_parser_id id, common_peg_parse_context & ctx, size_t start) const;
void resolve_refs();
void build_grammar(const common_grammar_builder & builder, bool lazy = false) const;
std::string dump(common_peg_parser_id id) const;
nlohmann::json to_json() const;
static common_peg_arena from_json(const nlohmann::json & j);
std::string save() const;
void load(const std::string & data);
friend class common_peg_parser_builder;
private:
common_peg_parser_id add_parser(common_peg_parser_variant parser);
void add_rule(const std::string & name, common_peg_parser_id id);
common_peg_parser_id resolve_ref(common_peg_parser_id id);
};
class common_peg_parser_builder {
common_peg_arena arena_;
common_peg_parser wrap(common_peg_parser_id id) { return common_peg_parser(id, *this); }
common_peg_parser add(const common_peg_parser_variant & p) { return wrap(arena_.add_parser(p)); }
public:
common_peg_parser_builder();
// Match nothing, always succeed.
// S -> ε
common_peg_parser eps() { return add(common_peg_epsilon_parser{}); }
// Matches the start of the input.
// S -> ^
common_peg_parser start() { return add(common_peg_start_parser{}); }
// Matches the end of the input.
// S -> $
common_peg_parser end() { return add(common_peg_end_parser{}); }
// Matches an exact literal string.
// S -> "hello"
common_peg_parser literal(const std::string & literal) { return add(common_peg_literal_parser{literal}); }
// Matches a sequence of parsers in order, all must succeed.
// S -> A B C
common_peg_parser sequence() { return add(common_peg_sequence_parser{}); }
common_peg_parser sequence(const std::vector<common_peg_parser_id> & parsers);
common_peg_parser sequence(const std::vector<common_peg_parser> & parsers);
common_peg_parser sequence(std::initializer_list<common_peg_parser> parsers);
// Matches the first parser that succeeds from a list of alternatives.
// S -> A | B | C
common_peg_parser choice() { return add(common_peg_choice_parser{}); }
common_peg_parser choice(const std::vector<common_peg_parser_id> & parsers);
common_peg_parser choice(const std::vector<common_peg_parser> & parsers);
common_peg_parser choice(std::initializer_list<common_peg_parser> parsers);
// Matches one or more repetitions of a parser.
// S -> A+
common_peg_parser one_or_more(const common_peg_parser & p) { return repeat(p, 1, -1); }
// Matches zero or more repetitions of a parser, always succeeds.
// S -> A*
common_peg_parser zero_or_more(const common_peg_parser & p) { return repeat(p, 0, -1); }
// Matches zero or one occurrence of a parser, always succeeds.
// S -> A?
common_peg_parser optional(const common_peg_parser & p) { return repeat(p, 0, 1); }
// Positive lookahead: succeeds if child parser succeeds, consumes no input.
// S -> &A
common_peg_parser peek(const common_peg_parser & p) { return add(common_peg_and_parser{p}); }
// Negative lookahead: succeeds if child parser fails, consumes no input.
// S -> !A
common_peg_parser negate(const common_peg_parser & p) { return add(common_peg_not_parser{p}); }
// Matches any single character.
// S -> .
common_peg_parser any() { return add(common_peg_any_parser{}); }
// Matches between min and max repetitions of characters from a character class.
// S -> [a-z]{m,n}
//
// Use -1 for max to represent unbounded repetition (equivalent to {m,})
common_peg_parser chars(const std::string & classes, int min = 1, int max = -1);
// Creates a lightweight reference to a named rule (resolved during build()).
// Use this for forward references in recursive grammars.
// expr_ref -> expr
common_peg_parser ref(const std::string & name) { return add(common_peg_ref_parser{name}); }
// Matches zero or more whitespace characters (space, tab, newline).
// S -> [ \t\n]*
common_peg_parser space() { return add(common_peg_space_parser{}); }
// Matches all characters until a delimiter is found (delimiter not consumed).
// S -> (!delim .)*
common_peg_parser until(const std::string & delimiter) { return add(common_peg_until_parser{{delimiter}}); }
// Matches all characters until one of the delimiters in the list is found (delimiter not consumed).
// S -> (!delim .)*
common_peg_parser until_one_of(const std::vector<std::string> & delimiters) { return add(common_peg_until_parser{delimiters}); }
// Matches everything
// S -> .*
common_peg_parser rest() { return until_one_of({}); }
// Matches between min and max repetitions of a parser (inclusive).
// S -> A{m,n}
// Use -1 for max to represent unbounded repetition (equivalent to {m,})
common_peg_parser repeat(const common_peg_parser & p, int min, int max) { return add(common_peg_repetition_parser{p, min,max}); }
// Matches exactly n repetitions of a parser.
// S -> A{n}
common_peg_parser repeat(const common_peg_parser & p, int n) { return repeat(p, n, n); }
// Creates a complete JSON parser supporting objects, arrays, strings, numbers, booleans, and null.
// value -> object | array | string | number | true | false | null
common_peg_parser json();
common_peg_parser json_object();
common_peg_parser json_string();
common_peg_parser json_array();
common_peg_parser json_number();
common_peg_parser json_bool();
common_peg_parser json_null();
// Matches JSON string content without the surrounding quotes.
// Useful for extracting content within a JSON string.
common_peg_parser json_string_content();
// Matches a JSON object member with a key and associated parser as the
// value.
common_peg_parser json_member(const std::string & key, const common_peg_parser & p);
// Wraps a parser with JSON schema metadata for grammar generation.
// Used internally to convert JSON schemas to GBNF grammar rules.
common_peg_parser schema(const common_peg_parser & p, const std::string & name, const nlohmann::ordered_json & schema, bool raw = false);
// Creates a named rule, stores it in the grammar, and returns a ref.
// If trigger=true, marks this rule as an entry point for lazy grammar generation.
// auto json = p.rule("json", json_obj | json_arr | ...)
common_peg_parser rule(const std::string & name, const common_peg_parser & p, bool trigger = false);
// Creates a named rule using a builder function, and returns a ref.
// If trigger=true, marks this rule as an entry point for lazy grammar generation.
// auto json = p.rule("json", [&]() { return json_object() | json_array() | ... })
common_peg_parser rule(const std::string & name, const std::function<common_peg_parser()> & builder, bool trigger = false);
// Creates a trigger rule. When generating a lazy grammar from the parser,
// only trigger rules and descendents are emitted.
common_peg_parser trigger_rule(const std::string & name, const common_peg_parser & p) { return rule(name, p, true); }
common_peg_parser trigger_rule(const std::string & name, const std::function<common_peg_parser()> & builder) { return rule(name, builder, true); }
// Creates an atomic parser. Atomic parsers do not create an AST node if
// the child results in a partial parse, i.e. NEEDS_MORE_INPUT. This is
// intended for situations where partial output is undesirable.
common_peg_parser atomic(const common_peg_parser & p) { return add(common_peg_atomic_parser{p}); }
// Tags create nodes in the generated AST for semantic purposes.
// Unlike rules, you can tag multiple nodes with the same tag.
common_peg_parser tag(const std::string & tag, const common_peg_parser & p) { return add(common_peg_tag_parser{p.id(), tag}); }
void set_root(const common_peg_parser & p);
common_peg_arena build();
};
// Helper function for building parsers
common_peg_arena build_peg_parser(const std::function<common_peg_parser(common_peg_parser_builder & builder)> & fn);

206
common/preset.cpp Normal file
View File

@ -0,0 +1,206 @@
#include "arg.h"
#include "preset.h"
#include "peg-parser.h"
#include "log.h"
#include <fstream>
#include <sstream>
#include <filesystem>
static std::string rm_leading_dashes(const std::string & str) {
size_t pos = 0;
while (pos < str.size() && str[pos] == '-') {
++pos;
}
return str.substr(pos);
}
std::vector<std::string> common_preset::to_args() const {
std::vector<std::string> args;
for (const auto & [opt, value] : options) {
args.push_back(opt.args.back()); // use the last arg as the main arg
if (opt.value_hint == nullptr && opt.value_hint_2 == nullptr) {
// flag option, no value
if (common_arg_utils::is_falsey(value)) {
// use negative arg if available
if (!opt.args_neg.empty()) {
args.back() = opt.args_neg.back();
} else {
// otherwise, skip the flag
// TODO: maybe throw an error instead?
args.pop_back();
}
}
}
if (opt.value_hint != nullptr) {
// single value
args.push_back(value);
}
if (opt.value_hint != nullptr && opt.value_hint_2 != nullptr) {
throw std::runtime_error(string_format(
"common_preset::to_args(): option '%s' has two values, which is not supported yet",
opt.args.back()
));
}
}
return args;
}
std::string common_preset::to_ini() const {
std::ostringstream ss;
ss << "[" << name << "]\n";
for (const auto & [opt, value] : options) {
auto espaced_value = value;
string_replace_all(espaced_value, "\n", "\\\n");
ss << rm_leading_dashes(opt.args.back()) << " = ";
ss << espaced_value << "\n";
}
ss << "\n";
return ss.str();
}
static std::map<std::string, std::map<std::string, std::string>> parse_ini_from_file(const std::string & path) {
std::map<std::string, std::map<std::string, std::string>> parsed;
if (!std::filesystem::exists(path)) {
throw std::runtime_error("preset file does not exist: " + path);
}
std::ifstream file(path);
if (!file.good()) {
throw std::runtime_error("failed to open server preset file: " + path);
}
std::string contents((std::istreambuf_iterator<char>(file)), std::istreambuf_iterator<char>());
static const auto parser = build_peg_parser([](auto & p) {
// newline ::= "\r\n" / "\n" / "\r"
auto newline = p.rule("newline", p.literal("\r\n") | p.literal("\n") | p.literal("\r"));
// ws ::= [ \t]*
auto ws = p.rule("ws", p.chars("[ \t]", 0, -1));
// comment ::= [;#] (!newline .)*
auto comment = p.rule("comment", p.chars("[;#]", 1, 1) + p.zero_or_more(p.negate(newline) + p.any()));
// eol ::= ws comment? (newline / EOF)
auto eol = p.rule("eol", ws + p.optional(comment) + (newline | p.end()));
// ident ::= [a-zA-Z_] [a-zA-Z0-9_.-]*
auto ident = p.rule("ident", p.chars("[a-zA-Z_]", 1, 1) + p.chars("[a-zA-Z0-9_.-]", 0, -1));
// value ::= (!eol-start .)*
auto eol_start = p.rule("eol-start", ws + (p.chars("[;#]", 1, 1) | newline | p.end()));
auto value = p.rule("value", p.zero_or_more(p.negate(eol_start) + p.any()));
// header-line ::= "[" ws ident ws "]" eol
auto header_line = p.rule("header-line", "[" + ws + p.tag("section-name", p.chars("[^]]")) + ws + "]" + eol);
// kv-line ::= ident ws "=" ws value eol
auto kv_line = p.rule("kv-line", p.tag("key", ident) + ws + "=" + ws + p.tag("value", value) + eol);
// comment-line ::= ws comment (newline / EOF)
auto comment_line = p.rule("comment-line", ws + comment + (newline | p.end()));
// blank-line ::= ws (newline / EOF)
auto blank_line = p.rule("blank-line", ws + (newline | p.end()));
// line ::= header-line / kv-line / comment-line / blank-line
auto line = p.rule("line", header_line | kv_line | comment_line | blank_line);
// ini ::= line* EOF
auto ini = p.rule("ini", p.zero_or_more(line) + p.end());
return ini;
});
common_peg_parse_context ctx(contents);
const auto result = parser.parse(ctx);
if (!result.success()) {
throw std::runtime_error("failed to parse server config file: " + path);
}
std::string current_section = COMMON_PRESET_DEFAULT_NAME;
std::string current_key;
ctx.ast.visit(result, [&](const auto & node) {
if (node.tag == "section-name") {
const std::string section = std::string(node.text);
current_section = section;
parsed[current_section] = {};
} else if (node.tag == "key") {
const std::string key = std::string(node.text);
current_key = key;
} else if (node.tag == "value" && !current_key.empty() && !current_section.empty()) {
parsed[current_section][current_key] = std::string(node.text);
current_key.clear();
}
});
return parsed;
}
static std::map<std::string, common_arg> get_map_key_opt(common_params_context & ctx_params) {
std::map<std::string, common_arg> mapping;
for (const auto & opt : ctx_params.options) {
for (const auto & env : opt.get_env()) {
mapping[env] = opt;
}
for (const auto & arg : opt.get_args()) {
mapping[rm_leading_dashes(arg)] = opt;
}
}
return mapping;
}
static bool is_bool_arg(const common_arg & arg) {
return !arg.args_neg.empty();
}
static std::string parse_bool_arg(const common_arg & arg, const std::string & key, const std::string & value) {
// if this is a negated arg, we need to reverse the value
for (const auto & neg_arg : arg.args_neg) {
if (rm_leading_dashes(neg_arg) == key) {
return common_arg_utils::is_truthy(value) ? "false" : "true";
}
}
// otherwise, not negated
return value;
}
common_presets common_presets_load(const std::string & path, common_params_context & ctx_params) {
common_presets out;
auto key_to_opt = get_map_key_opt(ctx_params);
auto ini_data = parse_ini_from_file(path);
for (auto section : ini_data) {
common_preset preset;
if (section.first.empty()) {
preset.name = COMMON_PRESET_DEFAULT_NAME;
} else {
preset.name = section.first;
}
LOG_DBG("loading preset: %s\n", preset.name.c_str());
for (const auto & [key, value] : section.second) {
LOG_DBG("option: %s = %s\n", key.c_str(), value.c_str());
if (key_to_opt.find(key) != key_to_opt.end()) {
auto & opt = key_to_opt[key];
if (is_bool_arg(opt)) {
preset.options[opt] = parse_bool_arg(opt, key, value);
} else {
preset.options[opt] = value;
}
LOG_DBG("accepted option: %s = %s\n", key.c_str(), preset.options[opt].c_str());
} else {
// TODO: maybe warn about unknown key?
}
}
out[preset.name] = preset;
}
return out;
}

32
common/preset.h Normal file
View File

@ -0,0 +1,32 @@
#pragma once
#include "common.h"
#include "arg.h"
#include <string>
#include <vector>
#include <map>
//
// INI preset parser and writer
//
constexpr const char * COMMON_PRESET_DEFAULT_NAME = "default";
struct common_preset {
std::string name;
// TODO: support repeated args in the future
std::map<common_arg, std::string> options;
// convert preset to CLI argument list
std::vector<std::string> to_args() const;
// convert preset to INI format string
std::string to_ini() const;
// TODO: maybe implement to_env() if needed
};
// interface for multiple presets in one file
using common_presets = std::map<std::string, common_preset>;
common_presets common_presets_load(const std::string & path, common_params_context & ctx_params);

183
common/sampling.cpp
View File

@ -104,9 +104,10 @@ struct ring_buffer {
struct common_sampler {
common_params_sampling params;
struct llama_sampler * grmr;
struct llama_sampler * chain;
bool grammar;
ring_buffer<llama_token> prev;
std::vector<llama_token_data> cur;
@ -116,7 +117,6 @@ struct common_sampler {
void reset() {
prev.clear();
llama_sampler_reset(grmr);
llama_sampler_reset(chain);
}
@ -167,10 +167,15 @@ struct common_sampler * common_sampler_init(const struct llama_model * model, co
lparams.no_perf = params.no_perf;
struct llama_sampler * grmr;
llama_sampler * chain = llama_sampler_chain_init(lparams);
bool grammar = false;
std::vector<llama_sampler *> samplers;
if (params.grammar.compare(0, 11, "%llguidance") == 0) {
#ifdef LLAMA_USE_LLGUIDANCE
grmr = llama_sampler_init_llg(vocab, "lark", params.grammar.c_str());
samplers.push_back(llama_sampler_init_llg(vocab, "lark", params.grammar.c_str()));
grammar = true;
#else
GGML_ABORT("llguidance (cmake -DLLAMA_LLGUIDANCE=ON) is not enabled");
#endif // LLAMA_USE_LLGUIDANCE
@ -217,30 +222,23 @@ struct common_sampler * common_sampler_init(const struct llama_model * model, co
trigger_patterns_c.push_back(regex.c_str());
}
grmr = params.grammar_lazy
? llama_sampler_init_grammar_lazy_patterns(vocab, params.grammar.c_str(), "root",
trigger_patterns_c.data(), trigger_patterns_c.size(),
trigger_tokens.data(), trigger_tokens.size())
: llama_sampler_init_grammar(vocab, params.grammar.c_str(), "root");
if (!grmr) {
return nullptr;
if (!params.grammar.empty()) {
if (params.grammar_lazy) {
samplers.push_back(
llama_sampler_init_grammar_lazy_patterns(vocab, params.grammar.c_str(), "root",
trigger_patterns_c.data(), trigger_patterns_c.size(),
trigger_tokens.data(), trigger_tokens.size()));
} else {
samplers.push_back(llama_sampler_init_grammar(vocab, params.grammar.c_str(), "root"));
}
grammar = true;
}
}
auto * result = new common_sampler {
/* .params = */ params,
/* .grmr = */ grmr,
/* .chain = */ llama_sampler_chain_init(lparams),
/* .prev = */ ring_buffer<llama_token>(std::max(32, params.n_prev)),
/* .cur = */ {},
/* .cur_p = */ {},
};
llama_sampler_chain_add(result->chain,
llama_sampler_init_logit_bias(
llama_vocab_n_tokens(vocab),
params.logit_bias.size(),
params.logit_bias.data()));
if (params.has_logit_bias()) {
samplers.push_back(llama_sampler_init_logit_bias(llama_vocab_n_tokens(vocab), params.logit_bias.size(), params.logit_bias.data()));
}
if (params.mirostat == 0) {
for (const auto & cnstr : params.samplers) {
@ -253,58 +251,70 @@ struct common_sampler * common_sampler_init(const struct llama_model * model, co
c_breakers.push_back(str.c_str());
}
llama_sampler_chain_add(result->chain, llama_sampler_init_dry (vocab, llama_model_n_ctx_train(model), params.dry_multiplier, params.dry_base, params.dry_allowed_length, params.dry_penalty_last_n, c_breakers.data(), c_breakers.size()));
samplers.push_back(llama_sampler_init_dry (vocab, llama_model_n_ctx_train(model), params.dry_multiplier, params.dry_base, params.dry_allowed_length, params.dry_penalty_last_n, c_breakers.data(), c_breakers.size()));
}
break;
case COMMON_SAMPLER_TYPE_TOP_K:
llama_sampler_chain_add(result->chain, llama_sampler_init_top_k (params.top_k));
samplers.push_back(llama_sampler_init_top_k (params.top_k));
break;
case COMMON_SAMPLER_TYPE_TOP_P:
llama_sampler_chain_add(result->chain, llama_sampler_init_top_p (params.top_p, params.min_keep));
samplers.push_back(llama_sampler_init_top_p (params.top_p, params.min_keep));
break;
case COMMON_SAMPLER_TYPE_TOP_N_SIGMA:
llama_sampler_chain_add(result->chain, llama_sampler_init_top_n_sigma (params.top_n_sigma));
samplers.push_back(llama_sampler_init_top_n_sigma(params.top_n_sigma));
break;
case COMMON_SAMPLER_TYPE_MIN_P:
llama_sampler_chain_add(result->chain, llama_sampler_init_min_p (params.min_p, params.min_keep));
samplers.push_back(llama_sampler_init_min_p (params.min_p, params.min_keep));
break;
case COMMON_SAMPLER_TYPE_XTC:
llama_sampler_chain_add(result->chain, llama_sampler_init_xtc (params.xtc_probability, params.xtc_threshold, params.min_keep, params.seed));
samplers.push_back(llama_sampler_init_xtc (params.xtc_probability, params.xtc_threshold, params.min_keep, params.seed));
break;
case COMMON_SAMPLER_TYPE_TYPICAL_P:
llama_sampler_chain_add(result->chain, llama_sampler_init_typical (params.typ_p, params.min_keep));
samplers.push_back(llama_sampler_init_typical (params.typ_p, params.min_keep));
break;
case COMMON_SAMPLER_TYPE_TEMPERATURE:
llama_sampler_chain_add(result->chain, llama_sampler_init_temp_ext (params.temp, params.dynatemp_range, params.dynatemp_exponent));
samplers.push_back(llama_sampler_init_temp_ext (params.temp, params.dynatemp_range, params.dynatemp_exponent));
break;
case COMMON_SAMPLER_TYPE_INFILL:
llama_sampler_chain_add(result->chain, llama_sampler_init_infill (vocab));
samplers.push_back(llama_sampler_init_infill (vocab));
break;
case COMMON_SAMPLER_TYPE_PENALTIES:
llama_sampler_chain_add(result->chain, llama_sampler_init_penalties (params.penalty_last_n, params.penalty_repeat, params.penalty_freq, params.penalty_present));
samplers.push_back(llama_sampler_init_penalties (params.penalty_last_n, params.penalty_repeat, params.penalty_freq, params.penalty_present));
break;
default:
GGML_ASSERT(false && "unknown sampler type");
}
}
llama_sampler_chain_add(result->chain, llama_sampler_init_dist(params.seed));
samplers.push_back(llama_sampler_init_dist(params.seed));
} else if (params.mirostat == 1) {
llama_sampler_chain_add(result->chain, llama_sampler_init_temp(params.temp));
llama_sampler_chain_add(result->chain, llama_sampler_init_mirostat(llama_vocab_n_tokens(vocab), params.seed, params.mirostat_tau, params.mirostat_eta, 100));
samplers.push_back(llama_sampler_init_temp(params.temp));
samplers.push_back(llama_sampler_init_mirostat(llama_vocab_n_tokens(vocab), params.seed, params.mirostat_tau, params.mirostat_eta, 100));
} else if (params.mirostat == 2) {
llama_sampler_chain_add(result->chain, llama_sampler_init_temp(params.temp));
llama_sampler_chain_add(result->chain, llama_sampler_init_mirostat_v2(params.seed, params.mirostat_tau, params.mirostat_eta));
samplers.push_back(llama_sampler_init_temp(params.temp));
samplers.push_back(llama_sampler_init_mirostat_v2(params.seed, params.mirostat_tau, params.mirostat_eta));
} else {
GGML_ASSERT(false && "unknown mirostat version");
}
for (auto * smpl : samplers) {
llama_sampler_chain_add(chain, smpl);
}
auto * result = new common_sampler {
/* .params = */ params,
/* .chain = */ chain,
/* .grammar = */ grammar,
/* .prev = */ ring_buffer<llama_token>(std::max(32, params.n_prev)),
/* .cur = */ {},
/* .cur_p = */ {},
};
return result;
}
void common_sampler_free(struct common_sampler * gsmpl) {
if (gsmpl) {
llama_sampler_free(gsmpl->grmr);
llama_sampler_free(gsmpl->chain);
delete gsmpl;
@ -314,11 +324,24 @@ void common_sampler_free(struct common_sampler * gsmpl) {
void common_sampler_accept(struct common_sampler * gsmpl, llama_token token, bool accept_grammar) {
const auto tm = gsmpl->tm();
if (accept_grammar) {
llama_sampler_accept(gsmpl->grmr, token);
}
if (gsmpl->grammar) {
const int n_smpl = llama_sampler_chain_n(gsmpl->chain);
llama_sampler_accept(gsmpl->chain, token);
for (int i = 0; i < n_smpl; i++) {
auto * smpl = llama_sampler_chain_get(gsmpl->chain, i);
// the grammar sampler is always the first one
if (i == 0) {
if (accept_grammar) {
llama_sampler_accept(smpl, token);
}
} else {
llama_sampler_accept(smpl, token);
}
}
} else {
llama_sampler_accept(gsmpl->chain, token);
}
gsmpl->prev.push_back(token);
}
@ -329,12 +352,12 @@ void common_sampler_reset(struct common_sampler * gsmpl) {
struct common_sampler * common_sampler_clone(common_sampler * gsmpl) {
return new common_sampler {
/* .params = */ gsmpl->params,
/* .grmr = */ llama_sampler_clone(gsmpl->grmr),
/* .chain = */ llama_sampler_clone(gsmpl->chain),
/* .prev = */ gsmpl->prev,
/* .cur = */ gsmpl->cur,
/* .cur_p = */ gsmpl->cur_p,
/* .params = */ gsmpl->params,
/* .chain = */ llama_sampler_clone(gsmpl->chain),
/* .grammar = */ gsmpl->grammar,
/* .prev = */ gsmpl->prev,
/* .cur = */ gsmpl->cur,
/* .cur_p = */ gsmpl->cur_p,
};
}
@ -383,58 +406,33 @@ void common_perf_print(const struct llama_context * ctx, const struct common_sam
}
}
llama_token common_sampler_sample(struct common_sampler * gsmpl, struct llama_context * ctx, int idx, bool grammar_first) {
struct llama_sampler * common_sampler_get(const struct common_sampler * gsmpl) {
return gsmpl->chain;
}
llama_token common_sampler_sample(struct common_sampler * gsmpl, struct llama_context * ctx, int idx) {
llama_synchronize(ctx);
// start measuring sampling time after the llama_context synchronization in order to not measure any ongoing async operations
const auto tm = gsmpl->tm();
gsmpl->set_logits(ctx, idx);
llama_token id = LLAMA_TOKEN_NULL;
auto & grmr = gsmpl->grmr;
auto & chain = gsmpl->chain;
auto & cur_p = gsmpl->cur_p; // initialized by set_logits
if (grammar_first) {
llama_sampler_apply(grmr, &cur_p);
}
gsmpl->set_logits(ctx, idx);
llama_sampler_apply(chain, &cur_p);
GGML_ASSERT(cur_p.selected != -1 && "no selected token during sampling - check your sampling configuration");
const llama_token id = cur_p.data[cur_p.selected].id;
id = cur_p.data[cur_p.selected].id;
if (grammar_first) {
return id;
}
// check if it the sampled token fits the grammar
{
llama_token_data single_token_data = { id, 1.0f, 0.0f };
llama_token_data_array single_token_data_array = { &single_token_data, 1, -1, false };
llama_sampler_apply(grmr, &single_token_data_array);
const bool is_valid = single_token_data_array.data[0].logit != -INFINITY;
if (is_valid) {
return id;
}
}
// resampling:
// if the token is not valid, sample again, but first apply the grammar sampler and then the sampling chain
gsmpl->set_logits(ctx, idx);
llama_sampler_apply(grmr, &cur_p);
llama_sampler_apply(chain, &cur_p);
GGML_ASSERT(cur_p.selected != -1 && "no selected token during re-sampling - check your sampling configuration");
return cur_p.data[cur_p.selected].id;
return id;
}
std::vector<llama_token> common_sampler_sample_and_accept_n(struct common_sampler * gsmpl, struct llama_context * ctx, const std::vector<int> & idxs, const llama_tokens & draft, bool grammar_first) {
std::vector<llama_token> common_sampler_sample_and_accept_n(struct common_sampler * gsmpl, struct llama_context * ctx, const std::vector<int> & idxs, const llama_tokens & draft) {
GGML_ASSERT(idxs.size() == draft.size() + 1 && "idxs.size() must be draft.size() + 1");
std::vector<llama_token> result;
@ -442,7 +440,7 @@ std::vector<llama_token> common_sampler_sample_and_accept_n(struct common_sample
size_t i = 0;
for (; i < draft.size(); i++) {
const llama_token id = common_sampler_sample(gsmpl, ctx, idxs[i], grammar_first);
const llama_token id = common_sampler_sample(gsmpl, ctx, idxs[i]);
common_sampler_accept(gsmpl, id, true);
@ -454,7 +452,7 @@ std::vector<llama_token> common_sampler_sample_and_accept_n(struct common_sample
}
if (i == draft.size()) {
const llama_token id = common_sampler_sample(gsmpl, ctx, idxs[i], grammar_first);
const llama_token id = common_sampler_sample(gsmpl, ctx, idxs[i]);
common_sampler_accept(gsmpl, id, true);
@ -464,13 +462,13 @@ std::vector<llama_token> common_sampler_sample_and_accept_n(struct common_sample
return result;
}
std::vector<llama_token> common_sampler_sample_and_accept_n(struct common_sampler * gsmpl, struct llama_context * ctx, const llama_tokens & draft, bool grammar_first) {
std::vector<llama_token> common_sampler_sample_and_accept_n(struct common_sampler * gsmpl, struct llama_context * ctx, const llama_tokens & draft) {
std::vector<int> idxs(draft.size() + 1);
for (size_t i = 0; i < idxs.size(); ++i) {
idxs[i] = i;
}
return common_sampler_sample_and_accept_n(gsmpl, ctx, idxs, draft, grammar_first);
return common_sampler_sample_and_accept_n(gsmpl, ctx, idxs, draft);
}
uint32_t common_sampler_get_seed(const struct common_sampler * gsmpl) {
@ -515,7 +513,8 @@ std::string common_sampler_print(const struct common_sampler * gsmpl) {
for (int i = 0; i < llama_sampler_chain_n(gsmpl->chain); i++) {
const auto * smpl = llama_sampler_chain_get(gsmpl->chain, i);
result += std::string("-> ") + llama_sampler_name(smpl) + " ";
result += std::string("-> ");
result += std::string(llama_sampler_name(smpl)) + " ";
}
return result;

17
common/sampling.h
View File

@ -48,6 +48,8 @@ struct common_sampler * common_sampler_clone (struct common_sampler * gsmpl);
// arguments can be nullptr to skip printing
void common_perf_print(const struct llama_context * ctx, const struct common_sampler * gsmpl);
struct llama_sampler * common_sampler_get(const struct common_sampler * gsmpl);
// extended sampling implementation:
//
// - set logits
@ -55,10 +57,7 @@ void common_perf_print(const struct llama_context * ctx, const struct common_sam
// - check if the token fits the grammar (if any)
// - if not: resample by first applying the grammar constraints and then sampling again (slower path)
//
// if grammar_first is true, the grammar is applied before the samplers (slower)
// useful in cases where all the resulting candidates (not just the sampled one) must fit the grammar
//
llama_token common_sampler_sample(struct common_sampler * gsmpl, struct llama_context * ctx, int idx, bool grammar_first = false);
llama_token common_sampler_sample(struct common_sampler * gsmpl, struct llama_context * ctx, int idx);
// generalized version of common_sampler_sample
//
@ -76,10 +75,10 @@ llama_token common_sampler_sample(struct common_sampler * gsmpl, struct llama_co
//
// returns at least 1 token, up to idxs.size()
//
std::vector<llama_token> common_sampler_sample_and_accept_n(struct common_sampler * gsmpl, struct llama_context * ctx, const std::vector<int> & idxs, const llama_tokens & draft, bool grammar_first = false);
std::vector<llama_token> common_sampler_sample_and_accept_n(struct common_sampler * gsmpl, struct llama_context * ctx, const std::vector<int> & idxs, const llama_tokens & draft);
// assume idxs == [ 0, 1, 2, ..., draft.size() ]
std::vector<llama_token> common_sampler_sample_and_accept_n(struct common_sampler * gsmpl, struct llama_context * ctx, const llama_tokens & draft, bool grammar_first = false);
std::vector<llama_token> common_sampler_sample_and_accept_n(struct common_sampler * gsmpl, struct llama_context * ctx, const llama_tokens & draft);
uint32_t common_sampler_get_seed(const struct common_sampler * gsmpl);
@ -107,3 +106,9 @@ std::vector<enum common_sampler_type> common_sampler_types_from_chars(const std:
llama_sampler * llama_sampler_init_llg(const llama_vocab * vocab,
const char * grammar_kind, const char * grammar_data);
struct common_sampler_deleter {
void operator()(common_sampler * s) { common_sampler_free(s); }
};
typedef std::unique_ptr<common_sampler, common_sampler_deleter> common_sampler_ptr;

2
common/speculative.cpp
View File

@ -315,7 +315,7 @@ llama_tokens common_speculative_gen_draft(
for (int i = 0; i < params.n_draft; ++i) {
common_batch_clear(batch);
common_sampler_sample(smpl, ctx_dft, 0, true);
common_sampler_sample(smpl, ctx_dft, 0);
const auto * cur_p = common_sampler_get_candidates(smpl, true);

64
common/unicode.cpp Normal file
View File

@ -0,0 +1,64 @@
#include "unicode.h"
// implementation adopted from src/unicode.cpp
size_t utf8_sequence_length(unsigned char first_byte) {
const size_t lookup[] = { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 3, 4 };
uint8_t highbits = static_cast<uint8_t>(first_byte) >> 4;
return lookup[highbits];
}
utf8_parse_result parse_utf8_codepoint(std::string_view input, size_t offset) {
if (offset >= input.size()) {
return utf8_parse_result(utf8_parse_result::INCOMPLETE);
}
// ASCII fast path
if (!(input[offset] & 0x80)) {
return utf8_parse_result(utf8_parse_result::SUCCESS, input[offset], 1);
}
// Invalid: continuation byte as first byte
if (!(input[offset] & 0x40)) {
return utf8_parse_result(utf8_parse_result::INVALID);
}
// 2-byte sequence
if (!(input[offset] & 0x20)) {
if (offset + 1 >= input.size()) {
return utf8_parse_result(utf8_parse_result::INCOMPLETE);
}
if ((input[offset + 1] & 0xc0) != 0x80) {
return utf8_parse_result(utf8_parse_result::INVALID);
}
auto result = ((input[offset] & 0x1f) << 6) | (input[offset + 1] & 0x3f);
return utf8_parse_result(utf8_parse_result::SUCCESS, result, 2);
}
// 3-byte sequence
if (!(input[offset] & 0x10)) {
if (offset + 2 >= input.size()) {
return utf8_parse_result(utf8_parse_result::INCOMPLETE);
}
if ((input[offset + 1] & 0xc0) != 0x80 || (input[offset + 2] & 0xc0) != 0x80) {
return utf8_parse_result(utf8_parse_result::INVALID);
}
auto result = ((input[offset] & 0x0f) << 12) | ((input[offset + 1] & 0x3f) << 6) | (input[offset + 2] & 0x3f);
return utf8_parse_result(utf8_parse_result::SUCCESS, result, 3);
}
// 4-byte sequence
if (!(input[offset] & 0x08)) {
if (offset + 3 >= input.size()) {
return utf8_parse_result(utf8_parse_result::INCOMPLETE);
}
if ((input[offset + 1] & 0xc0) != 0x80 || (input[offset + 2] & 0xc0) != 0x80 || (input[offset + 3] & 0xc0) != 0x80) {
return utf8_parse_result(utf8_parse_result::INVALID);
}
auto result = ((input[offset] & 0x07) << 18) | ((input[offset + 1] & 0x3f) << 12) | ((input[offset + 2] & 0x3f) << 6) | (input[offset + 3] & 0x3f);
return utf8_parse_result(utf8_parse_result::SUCCESS, result, 4);
}
// Invalid first byte
return utf8_parse_result(utf8_parse_result::INVALID);
}

22
common/unicode.h Normal file
View File

@ -0,0 +1,22 @@
#pragma once
#include <cstdint>
#include <string_view>
// UTF-8 parsing utilities for streaming-aware unicode support
struct utf8_parse_result {
uint32_t codepoint; // Decoded codepoint (only valid if status == SUCCESS)
size_t bytes_consumed; // How many bytes this codepoint uses (1-4)
enum status { SUCCESS, INCOMPLETE, INVALID } status;
utf8_parse_result(enum status s, uint32_t cp = 0, size_t bytes = 0)
: codepoint(cp), bytes_consumed(bytes), status(s) {}
};
// Determine the expected length of a UTF-8 sequence from its first byte
// Returns 0 for invalid first bytes
size_t utf8_sequence_length(unsigned char first_byte);
// Parse a single UTF-8 codepoint from input
utf8_parse_result parse_utf8_codepoint(std::string_view input, size_t offset);

770
convert_hf_to_gguf.py
View File

File diff suppressed because it is too large Load Diff

13
docs/backend/SYCL.md
View File

@ -42,6 +42,9 @@ The following releases are verified and recommended:
## News
- 2025.11
- Support malloc memory on device more than 4GB.
- 2025.2
- Optimize MUL_MAT Q4_0 on Intel GPU for all dGPUs and built-in GPUs since MTL. Increase the performance of LLM (llama-2-7b.Q4_0.gguf) 21%-87% on Intel GPUs (MTL, ARL-H, Arc, Flex, PVC).
|GPU|Base tokens/s|Increased tokens/s|Percent|
@ -789,6 +792,8 @@ use 1 SYCL GPUs: [0] with Max compute units:512
| GGML_SYCL_DISABLE_GRAPH | 0 or 1 (default) | Disable running computations through SYCL Graphs feature. Disabled by default because graph performance isn't yet better than non-graph performance. |
| GGML_SYCL_DISABLE_DNN | 0 (default) or 1 | Disable running computations through oneDNN and always use oneMKL. |
| ZES_ENABLE_SYSMAN | 0 (default) or 1 | Support to get free memory of GPU by sycl::aspect::ext_intel_free_memory.<br>Recommended to use when --split-mode = layer |
| UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS | 0 (default) or 1 | Support malloc device memory more than 4GB.|
## Known Issues
@ -835,6 +840,14 @@ use 1 SYCL GPUs: [0] with Max compute units:512
| The default context is too big. It leads to excessive memory usage.|Set `-c 8192` or a smaller value.|
| The model is too big and requires more memory than what is available.|Choose a smaller model or change to a smaller quantization, like Q5 -> Q4;<br>Alternatively, use more than one device to load model.|
- `ggml_backend_sycl_buffer_type_alloc_buffer: can't allocate 5000000000 Bytes of memory on device`
You need to enable to support 4GB memory malloc by:
```
export UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS=1
set UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS=1
```
### **GitHub contribution**:
Please add the `SYCL :` prefix/tag in issues/PRs titles to help the SYCL contributors to check/address them without delay.

258
docs/backend/ZenDNN.md Normal file
View File

@ -0,0 +1,258 @@
# llama.cpp for AMD ZenDNN
> [!WARNING]
> **Note:** ZenDNN is **not** the same as zDNN.
> - **ZenDNN** (this page): AMD's deep learning library for AMD EPYC CPUs
> - **zDNN**: IBM's Deep Neural Network acceleration library for IBM Z & LinuxONE Mainframes ([see zDNN documentation](zDNN.md))
- [Background](#background)
- [OS](#os)
- [Hardware](#hardware)
- [Supported Operations](#supported-operations)
- [DataType Supports](#datatype-supports)
- [Linux](#linux)
- [Environment Variable](#environment-variable)
- [Performance Optimization](#performance-optimization)
- [Known Issues](#known-issues)
- [TODO](#todo)
## Background
**ZenDNN** (Zen Deep Neural Network Library) is AMD's high-performance deep learning inference library optimized for AMD EPYC™ CPUs. It provides optimized implementations of key deep learning primitives and operations, delivering significant performance improvements for neural network workloads on AMD Zen-based processor architectures.
**Llama.cpp + ZenDNN**
The llama.cpp ZenDNN backend leverages AMD's optimized matrix multiplication primitives to accelerate inference on AMD CPUs. It utilizes ZenDNN's **LowOHA (Low Overhead Hardware Accelerated)** MatMul operator for efficient GEMM operations with minimal execution overhead, built-in weight caching, and direct access to backend libraries (AOCL BLIS, LibXSMM, OneDNN).
For more information about ZenDNN, visit: https://www.amd.com/en/developer/zendnn.html
## OS
| OS | Status | Verified |
|:-------:|:-------:|:----------------------------------------------:|
| Linux | Support | Ubuntu 20.04, 22.04, 24.04 |
For the latest list of supported operating systems, see the [ZenDNN Supported OS](https://github.com/amd/ZenDNN/blob/zendnnl/README.md#15-supported-os).
## Hardware
### AMD CPUs
**Recommended Processors**
ZenDNN is optimized for AMD EPYC™ processors and AMD Ryzen™ processors based on "Zen" microarchitecture and newer.
| CPU Family | Status | Notes |
|:-----------------------------:|:-------:|:----------------------------------:|
| AMD EPYC™ 9005 Series (Turin)| Support | 5th Gen - Zen 5 architecture |
| AMD EPYC™ 9004 Series (Genoa)| Support | 4th Gen - Zen 4 architecture |
| AMD EPYC™ 7003 Series (Milan)| Support | 3rd Gen - Zen 3 architecture |
| AMD Ryzen™ AI MAX (Strix Halo)| Support | High-performance mobile processors |
*Notes:*
- Best performance is achieved on AMD EPYC™ processors with high core counts (e.g., EPYC 9005 series).
- ZenDNN leverages AMD's advanced CPU features including AVX2 and AVX-512 instruction sets.
- For optimal performance, ensure your system has sufficient memory bandwidth.
## Supported Operations
The ZenDNN backend currently accelerates **matrix multiplication (MUL_MAT)** operations only. Other operations are handled by the standard CPU backend.
| Operation | Status | Notes |
|:-------------|:-------:|:----------------------------------------------:|
| MUL_MAT | ✓ | Accelerated via ZenDNN LowOHA MatMul |
*Note:* Since only MUL_MAT is accelerated, models will benefit most from ZenDNN when matrix multiplications dominate the computational workload (which is typical for transformer-based LLMs).
## DataType Supports
| DataType | Status | Notes |
|:----------------------:|:-------:|:---------------------------------------------:|
| FP32 | Support | Full precision floating point |
| BF16 | Support | BFloat16 (best performance on Zen 4/Zen 5) |
*Notes:*
- **BF16** provides best performance on Zen 4 and Zen 5 EPYC™ processors (Genoa, Turin).
## Linux
### I. Setup Environment
You have two options to set up ZenDNN:
#### Option 1: Automatic Download and Build (Recommended)
CMake will automatically download and build ZenDNN for you:
```sh
# Build llama.cpp - ZenDNN will be automatically downloaded and built
cmake -B build -DGGML_ZENDNN=ON -DCMAKE_BUILD_TYPE=Release
cmake --build build --config Release -j $(nproc)
```
No manual ZenDNN installation required. CMake will handle everything automatically.
#### Option 2: Use Custom ZenDNN Installation
If you want to build ZenDNN yourself or use a specific version:
**Step 1: Build ZenDNN from source**
```sh
# Clone ZenDNN repository
git clone https://github.com/amd/ZenDNN.git
cd ZenDNN
git checkout zendnnl
# Build and install (requires CMake >= 3.25)
mkdir build && cd build
cmake ..
cmake --build . --target all
```
Default installation path: `ZenDNN/build/install`
**For detailed build instructions**, refer to the [ZenDNN README](https://github.com/amd/ZenDNN/blob/zendnnl/README.md).
**Step 2: Build llama.cpp with custom ZenDNN path**
```sh
# Using environment variable
export ZENDNN_ROOT=/path/to/ZenDNN/build/install
cmake -B build -DGGML_ZENDNN=ON -DCMAKE_BUILD_TYPE=Release
cmake --build build --config Release -j $(nproc)
# OR specify path directly in CMake
cmake -B build -DGGML_ZENDNN=ON -DZENDNN_ROOT=/path/to/ZenDNN/build/install -DCMAKE_BUILD_TYPE=Release
cmake --build build --config Release -j $(nproc)
```
### II. Run the Server
#### 1. Download Model
Download LLaMA 3.1 8B Instruct BF16 model:
```sh
# Download from Hugging Face
huggingface-cli download meta-llama/Llama-3.1-8B-Instruct-GGUF --local-dir models/
```
#### 2. Start Server
Run llama.cpp server with ZenDNN acceleration:
```sh
# Set optimal configuration
export OMP_NUM_THREADS=64 # Adjust to your CPU core count
export ZENDNNL_MATMUL_ALGO=2 # Blocked AOCL BLIS for best performance
# Start server
./build/bin/llama-server \
-m models/Llama-3.1-8B-Instruct.BF16.gguf \
--host 0.0.0.0 \
--port 8080 \
-t 64
```
Access the server at `http://localhost:8080`.
**Performance tips**:
- Set `OMP_NUM_THREADS` to match your physical core count
- Use `ZENDNNL_MATMUL_ALGO=2` for optimal performance
- For NUMA systems: `numactl --cpunodebind=0 --membind=0 ./build/bin/llama-server ...`
## Environment Variable
### Build Time
| Name | Value | Function |
|--------------------|---------------------------------------|---------------------------------------------|
| GGML_ZENDNN | ON/OFF | Enable ZenDNN backend support |
| ZENDNN_ROOT | Path to ZenDNN installation | Set ZenDNN installation directory |
| GGML_OPENMP | ON/OFF (recommended: ON) | Enable OpenMP for multi-threading |
### Runtime
| Name | Value | Function |
|-------------------------|--------------------------|-------------------------------------------------------------------|
| OMP_NUM_THREADS | Number (e.g., 64) | Set number of OpenMP threads (recommended: physical core count) |
| ZENDNNL_MATMUL_ALGO | 0-5 | Select MatMul backend algorithm (see Performance Optimization) |
| ZENDNNL_PROFILE_LOG_LEVEL | 0-4 | Profiling log level (0=disabled, 4=verbose) |
| ZENDNNL_ENABLE_PROFILER | 0 or 1 | Enable detailed profiling (1=enabled) |
| ZENDNNL_API_LOG_LEVEL | 0-4 | API log level (0=disabled, 4=verbose) |
**Example**:
```sh
export OMP_NUM_THREADS=64
export ZENDNNL_MATMUL_ALGO=2 # Use Blocked AOCL BLIS for best performance
./build/bin/llama-cli -m models/llama-2-7b.Q4_0.gguf -p "Test" -n 100
```
## Performance Optimization
### MatMul Algorithm Selection
ZenDNN's LowOHA MatMul supports multiple backend algorithms. For **best performance**, use the **Blocked AOCL BLIS** algorithm:
```sh
export ZENDNNL_MATMUL_ALGO=2 # Blocked AOCL BLIS (recommended)
```
**Available algorithms**:
| Value | Algorithm | Description |
|:-----:|:-----------------------|:----------------------------------------------|
| 0 | Dynamic Dispatch | Automatic backend selection (default) |
| 1 | AOCL BLIS | AOCL BLIS backend |
| 2 | AOCL BLIS Blocked | **Blocked AOCL BLIS (recommended)** |
| 3 | OneDNN | OneDNN backend |
| 4 | OneDNN Blocked | Blocked OneDNN |
| 5 | LibXSMM | LibXSMM backend |
### Profiling and Debugging
For detailed profiling and logging options, refer to the [ZenDNN Logging Documentation](https://github.com/amd/ZenDNN/blob/zendnnl/docs/logging.md).
## Known Issues
- **Limited operation support**: Currently only matrix multiplication (MUL_MAT) is accelerated via ZenDNN. Other operations fall back to the standard CPU backend.
- **BF16 support**: BF16 operations require AMD Zen 4 or Zen 5 architecture (EPYC 9004/9005 series). On older CPUs, operations will use FP32.
- **NUMA awareness**: For multi-socket systems, manual NUMA binding may be required for optimal performance.
## Q&A
**Q: How do I verify that ZenDNN backend is being used?**
A: Check the log output when running llama.cpp. You should see messages indicating the ZenDNN backend is initialized. You can also check the backend name in the output.
**Q: What performance improvement can I expect?**
A: Performance gains vary depending on the model size, batch size, and CPU architecture. On AMD EPYC processors, you can typically expect 1.1x-2x speedup compared to standard CPU inference for matrix multiplication operations.
**Q: Can I use ZenDNN on non-AMD processors?**
A: ZenDNN is optimized specifically for AMD processors. While it may work on other x86-64 CPUs, performance benefits are only guaranteed on AMD Zen-based architectures.
**Q: Does ZenDNN support quantized models?**
A: Currently, ZenDNN primarily supports FP32 and BF16 data types. Quantized model support is not available at this time.
**Q: Why is my inference not faster with ZenDNN?**
A: Ensure:
1. You're using an AMD EPYC or Ryzen processor (Zen 2 or newer)
2. `OMP_NUM_THREADS` is set appropriately (physical core count)
3. `ZENDNNL_MATMUL_ALGO=2` is set for best performance (Blocked AOCL BLIS)
4. You're using a sufficiently large model (small models may not benefit as much)
5. Enable profiling to verify ZenDNN MatMul is being called
### **GitHub Contribution**:
Please add the **[ZenDNN]** prefix/tag in issues/PRs titles to help the ZenDNN-team check/address them without delay.
## TODO
- Expand operation support beyond MUL_MAT (attention operations, activations, etc.)

5
docs/backend/zDNN.md
View File

@ -1,5 +1,10 @@
# llama.cpp for IBM zDNN Accelerator
> [!WARNING]
> **Note:** zDNN is **not** the same as ZenDNN.
> - **zDNN** (this page): IBM's Deep Neural Network acceleration library for IBM Z & LinuxONE Mainframes
> - **ZenDNN**: AMD's deep learning library for AMD EPYC CPUs ([see ZenDNN documentation](ZenDNN.md))
## Background
IBM zDNN (Z Deep Neural Network) is a hardware acceleration library designed specifically to leverage the IBM NNPA (Neural Network Processor Assist) accelerator located within IBM Telum I and II processors. It provides significant performance improvements for neural network inference operations.

1
docs/build-riscv64-spacemit.md
View File

@ -19,6 +19,7 @@ cmake -B build \
-DGGML_RVV=ON \
-DGGML_RV_ZFH=ON \
-DGGML_RV_ZICBOP=ON \
-DGGML_RV_ZIHINTPAUSE=ON \
-DRISCV64_SPACEMIT_IME_SPEC=RISCV64_SPACEMIT_IME1 \
-DCMAKE_TOOLCHAIN_FILE=${PWD}/cmake/riscv64-spacemit-linux-gnu-gcc.cmake \
-DCMAKE_INSTALL_PREFIX=build/installed

43
docs/build.md
View File

@ -431,11 +431,22 @@ docker run -it --rm -v "$(pwd):/app:Z" --device /dev/dri/renderD128:/dev/dri/ren
### For Linux users:
#### Using the LunarG Vulkan SDK
First, follow the official LunarG instructions for the installation and setup of the Vulkan SDK in the [Getting Started with the Linux Tarball Vulkan SDK](https://vulkan.lunarg.com/doc/sdk/latest/linux/getting_started.html) guide.
> [!IMPORTANT]
> After completing the first step, ensure that you have used the `source` command on the `setup_env.sh` file inside of the Vulkan SDK in your current terminal session. Otherwise, the build won't work. Additionally, if you close out of your terminal, you must perform this step again if you intend to perform a build. However, there are ways to make this persistent. Refer to the Vulkan SDK guide linked in the first step for more information about any of this.
#### Using system packages
On Debian / Ubuntu, you can install the required dependencies using:
```sh
sudo apt-get install libvulkan-dev glslc
```
#### Common steps
Second, after verifying that you have followed all of the SDK installation/setup steps, use this command to make sure before proceeding:
```bash
vulkaninfo
@ -484,6 +495,38 @@ llama_new_context_with_model: CANN compute buffer size = 1260.81 MiB
For detailed info, such as model/device supports, CANN install, please refer to [llama.cpp for CANN](./backend/CANN.md).
## ZenDNN
ZenDNN provides optimized deep learning primitives for AMD EPYC™ CPUs. It accelerates matrix multiplication operations for inference workloads.
### Compilation
- Using `CMake` on Linux (automatic build):
```bash
cmake -B build -DGGML_ZENDNN=ON
cmake --build build --config Release
```
The first build will automatically download and build ZenDNN, which may take 5-10 minutes. Subsequent builds will be much faster.
- Using `CMake` with custom ZenDNN installation:
```bash
cmake -B build -DGGML_ZENDNN=ON -DZENDNN_ROOT=/path/to/zendnn/install
cmake --build build --config Release
```
### Testing
You can test with:
```bash
./build/bin/llama-cli -m PATH_TO_MODEL -p "Building a website can be done in 10 steps:" -n 50
```
For detailed information about hardware support, setup instructions, and performance optimization, refer to [llama.cpp for ZenDNN](./backend/ZenDNN.md).
## Arm® KleidiAI™
KleidiAI is a library of optimized microkernels for AI workloads, specifically designed for Arm CPUs. These microkernels enhance performance and can be enabled for use by the CPU backend.

3
docs/development/HOWTO-add-model.md
View File

@ -9,7 +9,8 @@ Adding a model requires few steps:
After following these steps, you can open PR.
Also, it is important to check that the examples and main ggml backends (CUDA, METAL, CPU) are working with the new architecture, especially:
- [main](/tools/main/)
- [cli](/tools/cli/)
- [completion](/tools/completion/)
- [imatrix](/tools/imatrix/)
- [quantize](/tools/quantize/)
- [server](/tools/server/)

288
docs/development/parsing.md Normal file
View File

@ -0,0 +1,288 @@
# Parsing Model Output
The `common` library contains a PEG parser implementation suitable for parsing
model output.
Types with the prefix `common_peg_*` are intended for general use and may have
applications beyond parsing model output, such as parsing user-provided regex
patterns.
Types with the prefix `common_chat_peg_*` are specialized helpers for model
output.
The parser features:
- Partial parsing of streaming input
- Built-in JSON parsers
- AST generation with semantics via "tagged" nodes
## Example
Below is a contrived example demonstrating how to use the PEG parser to parse
output from a model that emits arguments as JSON.
```cpp
auto parser = build_chat_peg_native_parser([&](common_chat_peg_native_builder & p) {
// Build a choice of all available tools
auto tool_choice = p.choice();
for (const auto & tool : tools) {
const auto & function = tool.at("function");
std::string name = function.at("name");
const auto & schema = function.at("parameters");
auto tool_name = p.json_member("name", "\"" + p.literal(name) + "\"");
auto tool_args = p.json_member("arguments", p.schema(p.json(), "tool-" + name + "-schema", schema));
tool_choice |= p.rule("tool-" + name, "{" << tool_name << "," << tool_args << "}");
}
// Define the tool call structure: <tool_call>[{tool}]</tool_call>
auto tool_call = p.trigger_rule("tool-call",
p.sequence({
p.literal("<tool_call>["),
tool_choice,
p.literal("]</tool_call>")
})
);
// Parser accepts content, optionally followed by a tool call
return p.sequence({
p.content(p.until("<tool_call>")),
p.optional(tool_call),
p.end()
});
});
```
For a more complete example, see `test_example_native()` in
[tests/test-chat-peg-parser.cpp](tests/test-chat-peg-parser.cpp).
## Parsers/Combinators
### Basic Matchers
- **`eps()`** - Matches nothing and always succeeds (epsilon/empty match)
- **`start()`** - Matches the start of input (anchor `^`)
- **`end()`** - Matches the end of input (anchor `$`)
- **`literal(string)`** - Matches an exact literal string
- **`any()`** - Matches any single character (`.`)
### Combinators
- **`sequence(...)`** - Matches parsers in order; all must succeed
- **`choice(...)`** - Matches the first parser that succeeds from alternatives (ordered choice)
- **`one_or_more(p)`** - Matches one or more repetitions (`+`)
- **`zero_or_more(p)`** - Matches zero or more repetitions (`*`)
- **`optional(p)`** - Matches zero or one occurrence (`?`)
- **`repeat(p, min, max)`** - Matches between min and max repetitions (use `-1` for unbounded)
- **`repeat(p, n)`** - Matches exactly n repetitions
### Lookahead
- **`peek(p)`** - Positive lookahead: succeeds if parser succeeds without consuming input (`&`)
- **`negate(p)`** - Negative lookahead: succeeds if parser fails without consuming input (`!`)
### Character Classes & Utilities
- **`chars(classes, min, max)`** - Matches repetitions of characters from a character class
- **`space()`** - Matches zero or more whitespace characters (space, tab, newline)
- **`until(delimiter)`** - Matches characters until delimiter is found (delimiter not consumed)
- **`until_one_of(delimiters)`** - Matches characters until any delimiter in the list is found
- **`rest()`** - Matches everything remaining (`.*`)
### JSON Parsers
- **`json()`** - Complete JSON parser (objects, arrays, strings, numbers, booleans, null)
- **`json_object()`** - JSON object parser
- **`json_array()`** - JSON array parser
- **`json_string()`** - JSON string parser
- **`json_number()`** - JSON number parser
- **`json_bool()`** - JSON boolean parser
- **`json_null()`** - JSON null parser
- **`json_string_content()`** - JSON string content without surrounding quotes
- **`json_member(key, p)`** - JSON object member with specific key and value parser
### Grammar Building
- **`ref(name)`** - Creates a lightweight reference to a named rule (for recursive grammars)
- **`rule(name, p, trigger)`** - Creates a named rule and returns a reference
- **`trigger_rule(name, p)`** - Creates a trigger rule (entry point for lazy grammar generation)
- **`schema(p, name, schema, raw)`** - Wraps parser with JSON schema metadata for grammar generation
### AST Control
- **`atomic(p)`** - Prevents AST node creation for partial parses
- **`tag(tag, p)`** - Creates AST nodes with semantic tags (multiple nodes can share tags)
## GBNF Grammar Generation
The PEG parser also acts as a convenient DSL for generating GBNF grammars, with
some exceptions.
```cpp
data.grammar = build_grammar([&](const common_grammar_builder & builder) {
foreach_function(params.tools, [&](const json & fn) {
builder.resolve_refs(fn.at("parameters"));
});
parser.build_grammar(builder, data.grammar_lazy);
});
```
The notable exception is the `negate(p)` lookahead parser, which cannot be
defined as a CFG grammar and therefore does not produce a rule. Its usage
should be limited and preferably hidden behind a `schema()` parser. In many
cases, `until(delimiter)` or `until_one_of(delimiters)` is a better choice.
Another limitation is that the PEG parser requires an unambiguous grammar. In
contrast, the `llama-grammar` implementation can support ambiguous grammars,
though they are difficult to parse.
### Lazy Grammars
During lazy grammar generation, only rules reachable from a `trigger_rule(p)`
are emitted in the grammar. All trigger rules are added as alternations in the
root rule. It is still necessary to define trigger patterns, as the parser has
no interaction with the grammar sampling.
### JSON Schema
The `schema(p, name, schema, raw)` parser will use the `json-schema-to-grammar`
implementation to generate the grammar instead of the underlying parser.
The `raw` option emits a grammar suitable for a raw string instead of a JSON
string. In other words, it won't be wrapped in quotes or require escaping
quotes. It should only be used when `type == "string"`.
The downside is that it can potentially lead to ambiguous grammars. For
example, if a user provides the pattern `^.*$`, the following grammar may be
generated:
```
root ::= "<arg>" .* "</arg>"
```
This creates an ambiguous grammar that cannot be parsed by the PEG parser. To
help mitigate this, if `.*` is found in the pattern, the grammar from the
underlying parser will be emitted instead.
## Common AST Shapes for Chat Parsing
Most model output can be placed in one of the following categories:
- Content only
- Tool calling with arguments emitted as a single JSON object
- Tool calling with arguments emitted as separate entities, either XML
(Qwen3-Coder, MiniMax M2) or pseudo-function calls (LFM2)
To provide broad coverage,
[`common/chat-peg-parser.h`](common/chat-peg-parser.h) contains builders and
mappers that help create parsers and visitors/extractors for these types. They
require parsers to tag nodes to conform to an AST "shape". This normalization
makes it easy to extract information and generalize parsing.
### Simple
The `common_chat_peg_builder` builds a `simple` parser that supports
content-only models with optional reasoning.
- **`reasoning(p)`** - Tag node for extracting `reasoning_content`
- **`content(p)`** - Tag node for extracting `content`
```cpp
build_chat_peg_parser([&](common_chat_peg_parser & p) {
return p.sequence({
p.optional("<think>" + p.reasoning(p.until("</think>")) + "</think>"),
p.content(p.until("<tool_call>")),
p.end()
});
});
```
Use `common_chat_peg_mapper` to extract the content. Note that this is already
done for you in `common_chat_peg_parser` when
`chat_format == COMMON_CHAT_FORMAT_PEG_SIMPLE`.
```cpp
auto result = parser.parse(ctx);
common_chat_msg msg;
auto mapper = common_chat_peg_mapper(msg);
mapper.from_ast(ctx.ast, result);
```
### Native
The `common_chat_peg_native_builder` builds a `native` parser suitable for
models that emit tool arguments as a direct JSON object.
- **`reasoning(p)`** - Tag node for `reasoning_content`
- **`content(p)`** - Tag node for `content`
- **`tool(p)`** - Tag entirety of a single tool call
- **`tool_open(p)`** - Tag start of a tool call
- **`tool_close(p)`** - Tag end of a tool call
- **`tool_id(p)`** - Tag the tool call ID (optional)
- **`tool_name(p)`** - Tag the tool name
- **`tool_args(p)`** - Tag the tool arguments
```cpp
build_chat_peg_native_parser([&](common_chat_peg_native_parser & p) {
auto get_weather_tool = p.tool(p.sequence({
p.tool_open(p.literal("{")),
p.json_member("name", "\"" + p.tool_name(p.literal("get_weather")) + "\""),
p.literal(","),
p.json_member("arguments", p.tool_args(p.json())),
p.tool_close(p.literal("}"))
}));
return p.sequence({
p.content(p.until("<tool_call>")),
p.literal("<tool_call>"),
get_weather_tool,
p.literal("</tool_call>"),
p.end()
});
});
```
### Constructed
The `common_chat_peg_constructed_builder` builds a `constructed` parser
suitable for models that emit tool arguments as separate entities, such as XML
tags.
- **`reasoning(p)`** - Tag node for `reasoning_content`
- **`content(p)`** - Tag node for `content`
- **`tool(p)`** - Tag entirety of a single tool call
- **`tool_open(p)`** - Tag start of a tool call
- **`tool_close(p)`** - Tag end of a tool call
- **`tool_name(p)`** - Tag the tool name
- **`tool_arg(p)`** - Tag a complete tool argument (name + value)
- **`tool_arg_open(p)`** - Tag start of a tool argument
- **`tool_arg_close(p)`** - Tag end of a tool argument
- **`tool_arg_name(p)`** - Tag the argument name
- **`tool_arg_string_value(p)`** - Tag string value for the argument
- **`tool_arg_json_value(p)`** - Tag JSON value for the argument
```cpp
build_chat_peg_constructed_parser([&](common_chat_peg_constructed_builder & p) {
auto location_arg = p.tool_arg(
p.tool_arg_open("<parameter name=\"" + p.tool_arg_name(p.literal("location")) + "\">"),
p.tool_arg_string_value(p.until("</parameter>")),
p.tool_arg_close(p.literal("</parameter>"))
);
auto get_weather_tool = p.tool(p.sequence({
p.tool_open("<function name=\"" + p.tool_name(p.literal("get_weather")) + "\">"),
location_arg,
p.tool_close(p.literal("</function>"))
}));
return p.sequence({
p.content(p.until("<tool_call>")),
p.literal("<tool_call>"),
get_weather_tool,
p.literal("</tool_call>"),
p.end()
});
});
```

6
docs/docker.md
View File

@ -56,7 +56,7 @@ docker run -v /path/to/models:/models ghcr.io/ggml-org/llama.cpp:light -m /model
or with a server image:
```bash
docker run -v /path/to/models:/models -p 8000:8000 ghcr.io/ggml-org/llama.cpp:server -m /models/7B/ggml-model-q4_0.gguf --port 8000 --host 0.0.0.0 -n 512
docker run -v /path/to/models:/models -p 8080:8080 ghcr.io/ggml-org/llama.cpp:server -m /models/7B/ggml-model-q4_0.gguf --port 8080 --host 0.0.0.0 -n 512
```
## Docker With CUDA
@ -91,7 +91,7 @@ After building locally, Usage is similar to the non-CUDA examples, but you'll ne
```bash
docker run --gpus all -v /path/to/models:/models local/llama.cpp:full-cuda --run -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512 --n-gpu-layers 1
docker run --gpus all -v /path/to/models:/models local/llama.cpp:light-cuda -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512 --n-gpu-layers 1
docker run --gpus all -v /path/to/models:/models local/llama.cpp:server-cuda -m /models/7B/ggml-model-q4_0.gguf --port 8000 --host 0.0.0.0 -n 512 --n-gpu-layers 1
docker run --gpus all -v /path/to/models:/models local/llama.cpp:server-cuda -m /models/7B/ggml-model-q4_0.gguf --port 8080 --host 0.0.0.0 -n 512 --n-gpu-layers 1
```
## Docker With MUSA
@ -125,5 +125,5 @@ After building locally, Usage is similar to the non-MUSA examples, but you'll ne
```bash
docker run -v /path/to/models:/models local/llama.cpp:full-musa --run -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512 --n-gpu-layers 1
docker run -v /path/to/models:/models local/llama.cpp:light-musa -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512 --n-gpu-layers 1
docker run -v /path/to/models:/models local/llama.cpp:server-musa -m /models/7B/ggml-model-q4_0.gguf --port 8000 --host 0.0.0.0 -n 512 --n-gpu-layers 1
docker run -v /path/to/models:/models local/llama.cpp:server-musa -m /models/7B/ggml-model-q4_0.gguf --port 8080 --host 0.0.0.0 -n 512 --n-gpu-layers 1
```

216
docs/ops.md
View File

@ -12,110 +12,112 @@ Legend:
- 🟡 Partially supported by this backend
- ❌ Not supported by this backend
| Operation | BLAS | CANN | CPU | CUDA | Metal | OpenCL | SYCL | Vulkan | zDNN |
|-----------|------|------|------|------|------|------|------|------|------|
| ABS | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ❌ |
| ACC | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
| ADD | ❌ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ✅ | ❌ |
| ADD1 | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ |
| ADD_ID | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
| ARANGE | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
| ARGMAX | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
| ARGSORT | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | 🟡 | ❌ |
| CEIL | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | 🟡 | 🟡 | ❌ |
| CLAMP | ❌ | ✅ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | 🟡 | ❌ |
| CONCAT | ❌ | ✅ | ✅ | 🟡 | ✅ | 🟡 | ✅ | ✅ | ❌ |
| CONT | ❌ | 🟡 | ✅ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ❌ |
| CONV_2D | ❌ | ❌ | ✅ | ✅ | ❌ | ✅ | ❌ | ✅ | ❌ |
| CONV_2D_DW | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
| CONV_3D | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| CONV_TRANSPOSE_1D | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
| CONV_TRANSPOSE_2D | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
| COS | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | 🟡 | 🟡 | ❌ |
| COUNT_EQUAL | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ |
| CPY | ❌ | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | ❌ |
| CROSS_ENTROPY_LOSS | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ |
| CROSS_ENTROPY_LOSS_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ |
| CUMSUM | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| DIAG_MASK_INF | ❌ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ✅ | ❌ |
| DIV | ❌ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ✅ | ❌ |
| DUP | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ❌ |
| ELU | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | ❌ | ❌ |
| EXP | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ❌ |
| EXPM1 | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | ❌ | ❌ | ❌ |
| FILL | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ✅ | ❌ |
| FLASH_ATTN_EXT | ❌ | 🟡 | ✅ | 🟡 | 🟡 | ❌ | ❌ | 🟡 | ❌ |
| FLOOR | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | 🟡 | 🟡 | ❌ |
| GATED_LINEAR_ATTN | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ | ❌ | ❌ |
| GEGLU | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ❌ |
| GEGLU_ERF | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ❌ |
| GEGLU_QUICK | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ❌ |
| GELU | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ❌ |
| GELU_ERF | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ❌ |
| GELU_QUICK | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ❌ |
| GET_ROWS | ❌ | 🟡 | ✅ | 🟡 | ✅ | 🟡 | 🟡 | 🟡 | ❌ |
| GET_ROWS_BACK | ❌ | ❌ | 🟡 | 🟡 | ❌ | ❌ | ❌ | ❌ | ❌ |
| GROUP_NORM | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
| GROUP_NORM_MUL_ADD | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| HARDSIGMOID | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ❌ |
| HARDSWISH | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ❌ |
| IM2COL | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | ✅ | ❌ |
| IM2COL_3D | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
| L2_NORM | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
| LEAKY_RELU | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | 🟡 | ❌ |
| LOG | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | 🟡 | ✅ | ❌ |
| MEAN | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
| MUL | ❌ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ✅ | ❌ |
| MUL_MAT | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 |
| MUL_MAT_ID | ❌ | 🟡 | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ❌ |
| NEG | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ❌ |
| NORM | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ❌ |
| NORM_MUL_ADD | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| OPT_STEP_ADAMW | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
| OPT_STEP_SGD | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
| OUT_PROD | 🟡 | ❌ | 🟡 | 🟡 | ❌ | ❌ | 🟡 | ❌ | ❌ |
| PAD | ❌ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ✅ | ❌ |
| PAD_REFLECT_1D | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ❌ | ❌ |
| POOL_2D | ❌ | 🟡 | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
| REGLU | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ❌ |
| RELU | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ❌ |
| REPEAT | ❌ | ✅ | ✅ | 🟡 | ✅ | 🟡 | ✅ | 🟡 | ❌ |
| REPEAT_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ |
| RMS_NORM | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | ✅ | ❌ |
| RMS_NORM_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ |
| RMS_NORM_MUL_ADD | ❌ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
| ROLL | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ |
| ROPE | ❌ | 🟡 | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
| ROPE_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
| ROUND | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | 🟡 | 🟡 | ❌ |
| RWKV_WKV6 | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
| RWKV_WKV7 | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
| SCALE | ❌ | 🟡 | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
| SET | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | 🟡 | ❌ | ❌ |
| SET_ROWS | ❌ | ❌ | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | ❌ |
| SGN | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | ❌ | ❌ |
| SIGMOID | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ❌ |
| SILU | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ❌ |
| SILU_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
| SIN | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | 🟡 | 🟡 | ❌ |
| SOFTCAP | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| SOFTPLUS | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | ❌ | 🟡 | ❌ |
| SOFT_MAX | ❌ | 🟡 | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
| SOFT_MAX_BACK | ❌ | ❌ | 🟡 | 🟡 | ❌ | ❌ | 🟡 | ✅ | ❌ |
| SOLVE_TRI | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| SQR | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | 🟡 | 🟡 | ❌ |
| SQRT | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | 🟡 | 🟡 | ❌ |
| SSM_CONV | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
| SSM_SCAN | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | 🟡 | ❌ |
| STEP | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ❌ |
| SUB | ❌ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ✅ | ❌ |
| SUM | ❌ | ✅ | ✅ | 🟡 | ❌ | ❌ | 🟡 | 🟡 | ❌ |
| SUM_ROWS | ❌ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ✅ | ❌ |
| SWIGLU | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ❌ |
| SWIGLU_OAI | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | 🟡 | ❌ |
| TANH | ❌ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ✅ | 🟡 | ❌ |
| TIMESTEP_EMBEDDING | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
| TRI | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| TRUNC | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | 🟡 | 🟡 | ❌ |
| UPSCALE | ❌ | 🟡 | ✅ | ✅ | 🟡 | ✅ | 🟡 | ✅ | ❌ |
| XIELU | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Operation | BLAS | CANN | CPU | CUDA | Metal | OpenCL | SYCL | Vulkan | WebGPU | ZenDNN | zDNN |
|-----------|------|------|------|------|------|------|------|------|------|------|------|
| ABS | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ✅ | ❌ | ❌ |
| ACC | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
| ADD | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
| ADD1 | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
| ADD_ID | ❌ | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ | ❌ |
| ARANGE | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
| ARGMAX | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
| ARGSORT | ❌ | ✅ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ❌ | ❌ | ❌ |
| CEIL | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | 🟡 | 🟡 | ❌ | ❌ | ❌ |
| CLAMP | ❌ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | 🟡 | ❌ | ❌ | ❌ |
| CONCAT | ❌ | ✅ | ✅ | 🟡 | ✅ | 🟡 | ✅ | ✅ | ❌ | ❌ | ❌ |
| CONT | ❌ | 🟡 | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | 🟡 | ❌ | ❌ |
| CONV_2D | ❌ | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ❌ | ❌ | ❌ |
| CONV_2D_DW | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
| CONV_3D | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| CONV_TRANSPOSE_1D | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
| CONV_TRANSPOSE_2D | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
| COS | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | ✅ | 🟡 | ❌ | ❌ | ❌ |
| COUNT_EQUAL | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
| CPY | ❌ | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | ❌ | ❌ |
| CROSS_ENTROPY_LOSS | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| CROSS_ENTROPY_LOSS_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| CUMSUM | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
| DIAG | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| DIAG_MASK_INF | ❌ | ✅ | ✅ | ✅ | ❌ | 🟡 | ✅ | ✅ | ❌ | ❌ | ❌ |
| DIV | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
| DUP | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | ✅ | ❌ | ❌ | ❌ |
| ELU | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | ❌ | ✅ | ❌ | ❌ |
| EXP | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ✅ | ❌ | ❌ |
| EXPM1 | ❌ | ❌ | ✅ | 🟡 | 🟡 | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| FILL | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
| FLASH_ATTN_EXT | ❌ | 🟡 | ✅ | 🟡 | 🟡 | 🟡 | ❌ | 🟡 | ❌ | ❌ | ❌ |
| FLOOR | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | 🟡 | 🟡 | ❌ | ❌ | ❌ |
| GATED_LINEAR_ATTN | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ |
| GEGLU | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ✅ | ❌ | ❌ |
| GEGLU_ERF | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ✅ | ❌ | ❌ |
| GEGLU_QUICK | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ✅ | ❌ | ❌ |
| GELU | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ✅ | ❌ | ❌ |
| GELU_ERF | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ✅ | ❌ | ❌ |
| GELU_QUICK | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ✅ | ❌ | ❌ |
| GET_ROWS | ❌ | 🟡 | ✅ | 🟡 | ✅ | 🟡 | 🟡 | 🟡 | 🟡 | ❌ | ❌ |
| GET_ROWS_BACK | ❌ | ❌ | 🟡 | 🟡 | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| GROUP_NORM | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ | ❌ |
| GROUP_NORM_MUL_ADD | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| HARDSIGMOID | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ✅ | ❌ | ❌ |
| HARDSWISH | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ✅ | ❌ | ❌ |
| IM2COL | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ | ❌ |
| IM2COL_3D | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
| L2_NORM | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
| LEAKY_RELU | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | ✅ | 🟡 | ❌ | ❌ | ❌ |
| LOG | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
| MEAN | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ | ❌ |
| MUL | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
| MUL_MAT | 🟡 | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 |
| MUL_MAT_ID | ❌ | 🟡 | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ❌ | ❌ | ❌ |
| NEG | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ✅ | ❌ | ❌ |
| NORM | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | 🟡 | ❌ | ❌ | ❌ |
| NORM_MUL_ADD | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| OPT_STEP_ADAMW | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
| OPT_STEP_SGD | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
| OUT_PROD | 🟡 | ❌ | 🟡 | 🟡 | ❌ | ❌ | 🟡 | ❌ | ❌ | ❌ | ❌ |
| PAD | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | 🟡 | ✅ | ❌ | ❌ | ❌ |
| PAD_REFLECT_1D | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ |
| POOL_2D | ❌ | 🟡 | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
| REGLU | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ✅ | ❌ | ❌ |
| RELU | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ✅ | ❌ | ❌ |
| REPEAT | ❌ | ✅ | ✅ | 🟡 | ✅ | 🟡 | ✅ | 🟡 | ❌ | ❌ | ❌ |
| REPEAT_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
| RMS_NORM | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
| RMS_NORM_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
| RMS_NORM_MUL_ADD | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| ROLL | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
| ROPE | ❌ | 🟡 | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
| ROPE_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
| ROUND | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | 🟡 | 🟡 | ❌ | ❌ | ❌ |
| RWKV_WKV6 | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
| RWKV_WKV7 | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
| SCALE | ❌ | 🟡 | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
| SET | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | 🟡 | ❌ | ❌ | ❌ | ❌ |
| SET_ROWS | ❌ | ❌ | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | ❌ | ❌ |
| SGN | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | ❌ | ✅ | ❌ | ❌ |
| SIGMOID | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ✅ | ❌ | ❌ |
| SILU | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ✅ | ❌ | ❌ |
| SILU_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
| SIN | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | ✅ | 🟡 | ❌ | ❌ | ❌ |
| SOFTCAP | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| SOFTPLUS | ❌ | ❌ | ✅ | 🟡 | 🟡 | ❌ | ❌ | 🟡 | ❌ | ❌ | ❌ |
| SOFT_MAX | ❌ | 🟡 | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
| SOFT_MAX_BACK | ❌ | ❌ | 🟡 | 🟡 | ❌ | ❌ | 🟡 | ✅ | ❌ | ❌ | ❌ |
| SOLVE_TRI | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | ❌ | 🟡 | ❌ | ❌ | ❌ |
| SQR | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ❌ | ❌ | ❌ |
| SQRT | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ❌ | ❌ | ❌ |
| SSM_CONV | ❌ | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ | ❌ |
| SSM_SCAN | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | 🟡 | ❌ | ❌ | ❌ |
| STEP | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ✅ | ❌ | ❌ |
| SUB | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
| SUM | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | 🟡 | 🟡 | ❌ | ❌ | ❌ |
| SUM_ROWS | ❌ | ✅ | ✅ | 🟡 | ✅ | 🟡 | 🟡 | ✅ | ❌ | ❌ | ❌ |
| SWIGLU | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ✅ | ❌ | ❌ |
| SWIGLU_OAI | ❌ | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | 🟡 | ✅ | ❌ | ❌ |
| TANH | ❌ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ✅ | 🟡 | ✅ | ❌ | ❌ |
| TIMESTEP_EMBEDDING | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ | ❌ |
| TOP_K | ❌ | ❌ | ✅ | ❌ | ✅ | ❌ | ❌ | 🟡 | ❌ | ❌ | ❌ |
| TRI | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
| TRUNC | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | 🟡 | 🟡 | ❌ | ❌ | ❌ |
| UPSCALE | ❌ | 🟡 | ✅ | ✅ | 🟡 | 🟡 | 🟡 | 🟡 | ❌ | ❌ | ❌ |
| XIELU | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ✅ | ❌ | ❌ |

19297
docs/ops/BLAS.csv
View File

File diff suppressed because it is too large Load Diff

496
docs/ops/CPU.csv
View File

@ -4964,6 +4964,7 @@
"CPU","CONV_TRANSPOSE_1D","ne_input=[2,1,1,1],ne_kernel=[3,1,1,1],s0=1,p0=0,d0=1","support","1","yes","CPU"
"CPU","CONV_TRANSPOSE_2D","ne_input=[3,2,3,1],ne_kernel=[2,2,1,3],stride=1","support","1","yes","CPU"
"CPU","CONV_TRANSPOSE_2D","ne_input=[10,10,9,1],ne_kernel=[3,3,1,9],stride=2","support","1","yes","CPU"
"CPU","CONV_TRANSPOSE_2D","ne_input=[129,63,35,1],ne_kernel=[3,3,48,35],stride=1","support","1","yes","CPU"
"CPU","COUNT_EQUAL","type=f32,ne=[4,500,1,1]","support","1","yes","CPU"
"CPU","COUNT_EQUAL","type=f32,ne=[4,5000,1,1]","support","1","yes","CPU"
"CPU","ARGMAX","type=f32,ne=[32,1,1,1]","support","1","yes","CPU"
@ -5419,17 +5420,45 @@
"CPU","CPY","type_src=f16,type_dst=f16,ne=[256,4,1,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","1","yes","CPU"
"CPU","CPY","type_src=f32,type_dst=f32,ne=[256,4,1,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","1","yes","CPU"
"CPU","CPY","type_src=bf16,type_dst=bf16,ne=[256,4,1,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","1","yes","CPU"
"CPU","CPY","type_src=i32,type_dst=i32,ne=[256,4,1,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","1","yes","CPU"
"CPU","CPY","type_src=i32,type_dst=i32,ne=[256,1,4,1],permute_src=[1,2,0,3],permute_dst=[0,0,0,0],_src_transpose=0","support","1","yes","CPU"
"CPU","CPY","type_src=f32,type_dst=f32,ne=[256,1,4,1],permute_src=[1,2,0,3],permute_dst=[0,0,0,0],_src_transpose=0","support","1","yes","CPU"
"CPU","CONT","type=f32,ne=[10,10,10,1]","support","1","yes","CPU"
"CPU","CONT","type=f32,ne=[2,1,1,1]","support","1","yes","CPU"
"CPU","CONT","type=f32,ne=[2,1,3,5]","support","1","yes","CPU"
"CPU","CONT","type=f32,ne=[2,3,5,7]","support","1","yes","CPU"
"CPU","CONT","type=f16,ne=[2,1,1,1]","support","1","yes","CPU"
"CPU","CONT","type=f16,ne=[2,1,3,5]","support","1","yes","CPU"
"CPU","CONT","type=f16,ne=[2,3,5,7]","support","1","yes","CPU"
"CPU","CONT","type=bf16,ne=[2,1,1,1]","support","1","yes","CPU"
"CPU","CONT","type=bf16,ne=[2,1,3,5]","support","1","yes","CPU"
"CPU","CONT","type=bf16,ne=[2,3,5,7]","support","1","yes","CPU"
"CPU","CONT","type=f32,ne=[2,1,1,1],use_view_slice=1","support","1","yes","CPU"
"CPU","CONT","type=f32,ne=[2,1,3,5],use_view_slice=1","support","1","yes","CPU"
"CPU","CONT","type=f32,ne=[2,3,5,7],use_view_slice=1","support","1","yes","CPU"
"CPU","CONT","type=f32,ne=[1,4,4,1],use_view_slice=1","support","1","yes","CPU"
"CPU","CONT","type=f32,ne=[1,8,17,1],use_view_slice=1","support","1","yes","CPU"
"CPU","CONT","type=f32,ne=[10,10,10,1],use_view_slice=1","support","1","yes","CPU"
"CPU","CONT","type=f32,ne=[2,1,1,1],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=f32,ne=[2,1,3,5],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=f32,ne=[2,3,5,7],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=f32,ne=[1,4,4,1],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=f32,ne=[1,8,17,1],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=f32,ne=[10,10,10,1],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=i32,ne=[2,1,1,1],use_view_slice=1","support","1","yes","CPU"
"CPU","CONT","type=i32,ne=[2,1,3,5],use_view_slice=1","support","1","yes","CPU"
"CPU","CONT","type=i32,ne=[2,3,5,7],use_view_slice=1","support","1","yes","CPU"
"CPU","CONT","type=i32,ne=[1,4,4,1],use_view_slice=1","support","1","yes","CPU"
"CPU","CONT","type=i32,ne=[1,8,17,1],use_view_slice=1","support","1","yes","CPU"
"CPU","CONT","type=i32,ne=[10,10,10,1],use_view_slice=1","support","1","yes","CPU"
"CPU","CONT","type=i32,ne=[2,1,1,1],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=i32,ne=[2,1,3,5],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=i32,ne=[2,3,5,7],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=i32,ne=[1,4,4,1],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=i32,ne=[1,8,17,1],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=i32,ne=[10,10,10,1],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=f16,ne=[2,1,1,1],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=f16,ne=[2,1,3,5],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=f16,ne=[2,3,5,7],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=f16,ne=[1,4,4,1],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=f16,ne=[1,8,17,1],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=f16,ne=[10,10,10,1],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=bf16,ne=[2,1,1,1],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=bf16,ne=[2,1,3,5],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=bf16,ne=[2,3,5,7],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=bf16,ne=[1,4,4,1],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=bf16,ne=[1,8,17,1],use_view_slice=0","support","1","yes","CPU"
"CPU","CONT","type=bf16,ne=[10,10,10,1],use_view_slice=0","support","1","yes","CPU"
"CPU","ADD","type=f16,ne=[1,1,8,1],nr=[1,1,1,1],nf=1","support","1","yes","CPU"
"CPU","SUB","type=f16,ne=[1,1,8,1],nr=[1,1,1,1],nf=1","support","1","yes","CPU"
"CPU","MUL","type=f16,ne=[1,1,8,1],nr=[1,1,1,1],nf=1","support","1","yes","CPU"
@ -5655,6 +5684,7 @@
"CPU","MUL","type=f32,ne=[64,262144,1,1],nr=[1,1,1,1],nf=1","support","1","yes","CPU"
"CPU","DIV","type=f32,ne=[64,262144,1,1],nr=[1,1,1,1],nf=1","support","1","yes","CPU"
"CPU","ADD1","type=f32,ne=[10,5,4,3]","support","1","yes","CPU"
"CPU","ADD1","type=f32,ne=[1024,1024,1,1]","support","1","yes","CPU"
"CPU","SCALE","type=f32,ne=[10,10,10,10],scale=2.000000,bias=0.000000,inplace=0","support","1","yes","CPU"
"CPU","SCALE","type=f32,ne=[10,10,10,10],scale=2.000000,bias=1.000000,inplace=0","support","1","yes","CPU"
"CPU","SCALE","type=f32,ne=[10,10,10,10],scale=2.000000,bias=1.000000,inplace=1","support","1","yes","CPU"
@ -8644,9 +8674,13 @@
"CPU","CLAMP","type=f16,ne=[7,1,5,3],min=-0.500000,max=0.500000","support","1","yes","CPU"
"CPU","LEAKY_RELU","type=f16,ne_a=[7,1,5,3],negative_slope=0.100000","support","1","yes","CPU"
"CPU","FLOOR","type=f16,ne=[7,1,5,3]","support","1","yes","CPU"
"CPU","FLOOR","type=f16,ne=[1024,1024,1,1]","support","1","yes","CPU"
"CPU","CEIL","type=f16,ne=[7,1,5,3]","support","1","yes","CPU"
"CPU","CEIL","type=f16,ne=[1024,1024,1,1]","support","1","yes","CPU"
"CPU","ROUND","type=f16,ne=[7,1,5,3]","support","1","yes","CPU"
"CPU","ROUND","type=f16,ne=[1024,1024,1,1]","support","1","yes","CPU"
"CPU","TRUNC","type=f16,ne=[7,1,5,3]","support","1","yes","CPU"
"CPU","TRUNC","type=f16,ne=[1024,1024,1,1]","support","1","yes","CPU"
"CPU","SQR","type=f32,ne=[10,5,4,3]","support","1","yes","CPU"
"CPU","SQRT","type=f32,ne=[10,3,3,2]","support","1","yes","CPU"
"CPU","LOG","type=f32,ne=[10,5,4,3]","support","1","yes","CPU"
@ -8666,9 +8700,13 @@
"CPU","CLAMP","type=f32,ne=[7,1,5,3],min=-0.500000,max=0.500000","support","1","yes","CPU"
"CPU","LEAKY_RELU","type=f32,ne_a=[7,1,5,3],negative_slope=0.100000","support","1","yes","CPU"
"CPU","FLOOR","type=f32,ne=[7,1,5,3]","support","1","yes","CPU"
"CPU","FLOOR","type=f32,ne=[1024,1024,1,1]","support","1","yes","CPU"
"CPU","CEIL","type=f32,ne=[7,1,5,3]","support","1","yes","CPU"
"CPU","CEIL","type=f32,ne=[1024,1024,1,1]","support","1","yes","CPU"
"CPU","ROUND","type=f32,ne=[7,1,5,3]","support","1","yes","CPU"
"CPU","ROUND","type=f32,ne=[1024,1024,1,1]","support","1","yes","CPU"
"CPU","TRUNC","type=f32,ne=[7,1,5,3]","support","1","yes","CPU"
"CPU","TRUNC","type=f32,ne=[1024,1024,1,1]","support","1","yes","CPU"
"CPU","DIAG_MASK_INF","type=f32,ne=[10,10,1,1],n_past=5","support","1","yes","CPU"
"CPU","DIAG_MASK_INF","type=f32,ne=[10,10,3,1],n_past=5","support","1","yes","CPU"
"CPU","DIAG_MASK_INF","type=f32,ne=[10,10,3,2],n_past=5","support","1","yes","CPU"
@ -9411,18 +9449,405 @@
"CPU","CONCAT","type=i32,ne_a=[11,12,13,14],ne_b_d=7,dim=2,v=3","support","1","yes","CPU"
"CPU","CONCAT","type=f32,ne_a=[11,12,13,14],ne_b_d=7,dim=3,v=3","support","1","yes","CPU"
"CPU","CONCAT","type=i32,ne_a=[11,12,13,14],ne_b_d=7,dim=3,v=3","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[3,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[4,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[7,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[8,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[15,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[16,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[31,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[32,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[63,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[64,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[127,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[128,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[255,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[256,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[511,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[512,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[1023,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[1024,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[2047,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[2048,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[4095,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[4096,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[8191,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[8192,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[16383,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[16384,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[32767,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[32768,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[65535,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[65536,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[131071,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[131072,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[262143,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[262144,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[524287,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[524288,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[1048575,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[1048576,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[16,10,10,10],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[60,10,10,10],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[1024,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[16384,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[1023,2,1,3],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[1024,2,1,3],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[1025,2,1,3],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[2047,2,1,3],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[2048,2,1,3],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[2049,2,1,3],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[2,8,8192,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[8,1,1,1],order=1","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[3,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[4,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[7,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[8,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[15,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[16,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[31,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[32,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[63,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[64,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[127,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[128,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[255,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[256,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[511,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[512,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[1023,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[1024,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[2047,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[2048,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[4095,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[4096,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[8191,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[8192,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[16383,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[16384,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[32767,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[32768,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[65535,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[65536,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[131071,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[131072,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[262143,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[262144,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[524287,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[524288,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[1048575,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[1048576,1,1,1],order=0","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[16,10,10,10],order=1","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[60,10,10,10],order=1","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[1024,1,1,1],order=1","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[16384,1,1,1],order=1","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[1023,2,1,3],order=1","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[1024,2,1,3],order=1","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[1025,2,1,3],order=1","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[2047,2,1,3],order=1","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[2048,2,1,3],order=1","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[2049,2,1,3],order=1","support","1","yes","CPU"
"CPU","ARGSORT","type=f32,ne=[2,8,8192,1],order=1","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[12,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[13,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[13,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[15,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[15,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[15,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[19,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[19,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[19,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[19,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[27,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[27,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[27,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[27,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[27,1,2,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[43,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[43,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[43,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[43,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[43,1,2,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[64,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[75,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[64,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[75,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[64,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[75,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[64,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[75,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[64,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[75,1,2,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[128,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[139,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[128,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[139,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[128,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[139,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[128,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[139,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[128,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[139,1,2,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[128,1,1,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[139,1,2,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[256,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[267,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[256,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[267,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[256,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[267,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[256,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[267,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[256,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[267,1,2,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[256,1,1,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[267,1,2,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[512,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[523,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[512,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[523,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[512,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[523,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[512,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[523,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[512,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[523,1,2,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[512,1,1,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[523,1,2,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[512,1,1,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[523,1,2,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1024,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1035,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1024,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1035,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1024,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1035,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1024,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1035,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1024,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1035,1,2,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1024,1,1,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1035,1,2,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1024,1,1,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1035,1,2,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1024,1,1,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1035,1,2,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2048,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2059,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2048,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2059,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2048,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2059,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2048,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2059,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2048,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2059,1,2,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2048,1,1,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2059,1,2,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2048,1,1,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2059,1,2,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2048,1,1,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2059,1,2,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4096,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4107,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4096,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4107,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4096,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4107,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4096,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4107,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4096,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4107,1,2,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4096,1,1,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4107,1,2,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4096,1,1,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4107,1,2,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4096,1,1,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[4107,1,2,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8192,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8203,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8192,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8203,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8192,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8203,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8192,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8203,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8192,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8203,1,2,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8192,1,1,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8203,1,2,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8192,1,1,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8203,1,2,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8192,1,1,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[8203,1,2,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16384,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16395,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16384,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16395,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16384,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16395,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16384,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16395,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16384,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16395,1,2,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16384,1,1,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16395,1,2,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16384,1,1,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16395,1,2,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16384,1,1,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16395,1,2,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16384,1,1,1],k=9999,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16395,1,2,1],k=9999,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32768,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32779,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32768,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32779,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32768,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32779,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32768,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32779,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32768,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32779,1,2,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32768,1,1,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32779,1,2,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32768,1,1,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32779,1,2,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32768,1,1,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32779,1,2,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32768,1,1,1],k=9999,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[32779,1,2,1],k=9999,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65536,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65547,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65536,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65547,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65536,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65547,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65536,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65547,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65536,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65547,1,2,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65536,1,1,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65547,1,2,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65536,1,1,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65547,1,2,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65536,1,1,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65547,1,2,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65536,1,1,1],k=9999,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[65547,1,2,1],k=9999,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131072,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131083,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131072,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131083,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131072,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131083,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131072,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131083,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131072,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131083,1,2,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131072,1,1,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131083,1,2,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131072,1,1,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131083,1,2,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131072,1,1,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131083,1,2,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131072,1,1,1],k=9999,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[131083,1,2,1],k=9999,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262144,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262155,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262144,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262155,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262144,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262155,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262144,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262155,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262144,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262155,1,2,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262144,1,1,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262155,1,2,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262144,1,1,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262155,1,2,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262144,1,1,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262155,1,2,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262144,1,1,1],k=9999,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[262155,1,2,1],k=9999,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524288,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524299,1,2,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524288,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524299,1,2,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524288,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524299,1,2,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524288,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524299,1,2,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524288,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524299,1,2,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524288,1,1,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524299,1,2,1],k=100,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524288,1,1,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524299,1,2,1],k=500,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524288,1,1,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524299,1,2,1],k=1023,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524288,1,1,1],k=9999,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[524299,1,2,1],k=9999,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16,10,10,10],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[60,10,10,10],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1023,2,1,3],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1024,2,1,3],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1025,2,1,3],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16384,1,1,1],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2047,2,1,3],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2048,2,1,3],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2049,2,1,3],k=1,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16,10,10,10],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[60,10,10,10],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1023,2,1,3],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1024,2,1,3],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1025,2,1,3],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16384,1,1,1],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2047,2,1,3],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2048,2,1,3],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2049,2,1,3],k=2,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16,10,10,10],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[60,10,10,10],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1023,2,1,3],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1024,2,1,3],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1025,2,1,3],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16384,1,1,1],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2047,2,1,3],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2048,2,1,3],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2049,2,1,3],k=3,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16,10,10,10],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[60,10,10,10],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1023,2,1,3],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1024,2,1,3],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1025,2,1,3],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16384,1,1,1],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2047,2,1,3],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2048,2,1,3],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2049,2,1,3],k=7,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16,10,10,10],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[60,10,10,10],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1023,2,1,3],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1024,2,1,3],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[1025,2,1,3],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[16384,1,1,1],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2047,2,1,3],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2048,2,1,3],k=15,ties=0","support","1","yes","CPU"
"CPU","TOP_K","type=f32,ne=[2049,2,1,3],k=15,ties=0","support","1","yes","CPU"
"CPU","UPSCALE","type=f32,ne=[512,512,3,2],scale_factor=2,mode=nearest,transpose=0","support","1","yes","CPU"
"CPU","UPSCALE","type=f32,ne=[512,512,3,2],scale_factor=2,mode=nearest,transpose=1","support","1","yes","CPU"
"CPU","UPSCALE","type=f32,ne=[2,5,7,11],ne_tgt=[5,7,11,13],mode=nearest,flags=none","support","1","yes","CPU"
@ -9435,6 +9860,10 @@
"CPU","UPSCALE","type=f32,ne=[512,512,3,2],scale_factor=2,mode=bicubic,transpose=1","support","1","yes","CPU"
"CPU","UPSCALE","type=f32,ne=[2,5,7,11],ne_tgt=[5,7,11,13],mode=bicubic,flags=none","support","1","yes","CPU"
"CPU","UPSCALE","type=f32,ne=[5,7,11,13],ne_tgt=[2,5,7,11],mode=bicubic,flags=none","support","1","yes","CPU"
"CPU","UPSCALE","type=f32,ne=[512,512,3,2],scale_factor=2,mode=513,transpose=0","support","1","yes","CPU"
"CPU","UPSCALE","type=f32,ne=[512,512,3,2],scale_factor=2,mode=513,transpose=1","support","1","yes","CPU"
"CPU","UPSCALE","type=f32,ne=[2,5,7,11],ne_tgt=[5,7,11,13],mode=bilinear,flags=none","support","1","yes","CPU"
"CPU","UPSCALE","type=f32,ne=[5,7,11,13],ne_tgt=[2,5,7,11],mode=bilinear,flags=none","support","1","yes","CPU"
"CPU","UPSCALE","type=f32,ne=[2,5,7,11],ne_tgt=[5,7,11,13],mode=bilinear,flags=align_corners","support","1","yes","CPU"
"CPU","UPSCALE","type=f32,ne=[1,4,3,2],ne_tgt=[2,8,3,2],mode=bilinear,flags=align_corners","support","1","yes","CPU"
"CPU","UPSCALE","type=f32,ne=[4,1,3,2],ne_tgt=[1,1,3,2],mode=bilinear,flags=align_corners","support","1","yes","CPU"
@ -9463,15 +9892,30 @@
"CPU","GROUP_NORM","type=f32,ne=[64,64,320,1],num_groups=32,eps=0.000001","support","1","yes","CPU"
"CPU","GROUP_NORM","type=f32,ne=[9,9,1280,1],num_groups=32,eps=0.000001","support","1","yes","CPU"
"CPU","ACC","type=f32,ne_a=[256,17,1,1],ne_b=[256,16,1,1]","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[512,512,1,1],pad_0=1,pad_1=1","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[512,512,3,1],lp0=1,rp0=1,lp1=1,rp1=1,lp2=1,rp2=1,lp3=1,rp3=1,v=0","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[512,512,1,1],pad_0=1,pad_1=1,circular=0","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[33,17,2,1],pad_0=4,pad_1=3,circular=1","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[512,512,3,1],lp0=1,rp0=1,lp1=1,rp1=1,lp2=1,rp2=1,lp3=1,rp3=1,v=0,circular=0","support","1","yes","CPU"
"CPU","PAD_REFLECT_1D","type=f32,ne_a=[512,34,2,1],pad_0=10,pad_1=9","support","1","yes","CPU"
"CPU","PAD_REFLECT_1D","type=f32,ne_a=[3000,384,4,1],pad_0=10,pad_1=9","support","1","yes","CPU"
"CPU","ROLL","shift0=3,shift1=-2,shift3=1,shift4=-1","support","1","yes","CPU"
"CPU","ARANGE","type=f32,start=0.000000,stop=10.000000,step=1.000000","support","1","yes","CPU"
"CPU","ARANGE","type=f32,start=0.000000,stop=1048576.000000,step=1.000000","support","1","yes","CPU"
"CPU","TIMESTEP_EMBEDDING","type=f32,ne_a=[2,1,1,1],dim=320,max_period=10000","support","1","yes","CPU"
"CPU","LEAKY_RELU","type=f32,ne_a=[10,5,4,3],negative_slope=0.100000","support","1","yes","CPU"
"CPU","CUMSUM","type=f32,ne=[10,5,4,3]","support","1","yes","CPU"
"CPU","CUMSUM","type=f32,ne=[127,5,4,3]","support","1","yes","CPU"
"CPU","CUMSUM","type=f32,ne=[128,5,4,3]","support","1","yes","CPU"
"CPU","CUMSUM","type=f32,ne=[128,128,4,4]","support","1","yes","CPU"
"CPU","CUMSUM","type=f32,ne=[255,5,4,3]","support","1","yes","CPU"
"CPU","CUMSUM","type=f32,ne=[256,5,4,3]","support","1","yes","CPU"
"CPU","CUMSUM","type=f32,ne=[511,5,4,3]","support","1","yes","CPU"
"CPU","CUMSUM","type=f32,ne=[512,5,4,3]","support","1","yes","CPU"
"CPU","CUMSUM","type=f32,ne=[1023,5,4,3]","support","1","yes","CPU"
"CPU","CUMSUM","type=f32,ne=[1024,5,4,3]","support","1","yes","CPU"
"CPU","CUMSUM","type=f32,ne=[2047,5,4,3]","support","1","yes","CPU"
"CPU","CUMSUM","type=f32,ne=[2048,5,4,3]","support","1","yes","CPU"
"CPU","CUMSUM","type=f32,ne=[242004,1,1,1]","support","1","yes","CPU"
"CPU","CUMSUM","type=f32,ne=[375960,1,1,1]","support","1","yes","CPU"
"CPU","XIELU","type=f32,ne=[10,5,4,3]","support","1","yes","CPU"
"CPU","TRI","type=f32,ne=[10,10,4,3],tri_type=3","support","1","yes","CPU"
"CPU","TRI","type=f32,ne=[10,10,4,3],tri_type=2","support","1","yes","CPU"
@ -9480,6 +9924,10 @@
"CPU","FILL","type=f32,ne=[10,10,4,3],c=0.000000","support","1","yes","CPU"
"CPU","FILL","type=f32,ne=[303,207,11,3],c=2.000000","support","1","yes","CPU"
"CPU","FILL","type=f32,ne=[800,600,4,4],c=-152.000000","support","1","yes","CPU"
"CPU","FILL","type=f32,ne=[2048,512,2,2],c=3.500000","support","1","yes","CPU"
"CPU","DIAG","type=f32,ne=[10,1,4,3]","support","1","yes","CPU"
"CPU","DIAG","type=f32,ne=[79,1,19,13]","support","1","yes","CPU"
"CPU","DIAG","type=f32,ne=[256,1,8,16]","support","1","yes","CPU"
"CPU","SOLVE_TRI","type=f32,ne_lhs=[10,10,4,3],ne_rhs=[3,10,4,3]","support","1","yes","CPU"
"CPU","SOLVE_TRI","type=f32,ne_lhs=[11,11,1,1],ne_rhs=[5,11,1,1]","support","1","yes","CPU"
"CPU","SOLVE_TRI","type=f32,ne_lhs=[17,17,2,4],ne_rhs=[9,17,2,4]","support","1","yes","CPU"
@ -9487,10 +9935,16 @@
"CPU","SOLVE_TRI","type=f32,ne_lhs=[42,42,5,2],ne_rhs=[10,42,5,2]","support","1","yes","CPU"
"CPU","SOLVE_TRI","type=f32,ne_lhs=[64,64,2,2],ne_rhs=[10,64,2,2]","support","1","yes","CPU"
"CPU","SOLVE_TRI","type=f32,ne_lhs=[100,100,4,4],ne_rhs=[41,100,4,4]","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[512,512,1,1],lp0=0,rp0=1,lp1=0,rp1=1,lp2=0,rp2=0,lp3=0,rp3=0,v=0","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[11,22,33,44],lp0=1,rp0=2,lp1=3,rp1=4,lp2=5,rp2=6,lp3=7,rp3=8,v=0","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[512,512,1,1],lp0=0,rp0=1,lp1=0,rp1=1,lp2=0,rp2=0,lp3=0,rp3=0,v=1","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[11,22,33,44],lp0=1,rp0=2,lp1=3,rp1=4,lp2=5,rp2=6,lp3=7,rp3=8,v=1","support","1","yes","CPU"
"CPU","SOLVE_TRI","type=f32,ne_lhs=[128,128,4,4],ne_rhs=[31,128,4,4]","support","1","yes","CPU"
"CPU","SOLVE_TRI","type=f32,ne_lhs=[64,64,4,4],ne_rhs=[300,64,4,4]","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[512,512,1,1],lp0=0,rp0=1,lp1=0,rp1=1,lp2=0,rp2=0,lp3=0,rp3=0,v=0,circular=0","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[11,22,33,44],lp0=1,rp0=2,lp1=3,rp1=4,lp2=5,rp2=6,lp3=7,rp3=8,v=0,circular=0","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[512,512,1,1],lp0=0,rp0=1,lp1=0,rp1=1,lp2=0,rp2=0,lp3=0,rp3=0,v=0,circular=1","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[11,22,33,44],lp0=1,rp0=2,lp1=3,rp1=4,lp2=5,rp2=6,lp3=7,rp3=8,v=0,circular=1","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[512,512,1,1],lp0=0,rp0=1,lp1=0,rp1=1,lp2=0,rp2=0,lp3=0,rp3=0,v=1,circular=0","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[11,22,33,44],lp0=1,rp0=2,lp1=3,rp1=4,lp2=5,rp2=6,lp3=7,rp3=8,v=1,circular=0","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[512,512,1,1],lp0=0,rp0=1,lp1=0,rp1=1,lp2=0,rp2=0,lp3=0,rp3=0,v=1,circular=1","support","1","yes","CPU"
"CPU","PAD","type=f32,ne_a=[11,22,33,44],lp0=1,rp0=2,lp1=3,rp1=4,lp2=5,rp2=6,lp3=7,rp3=8,v=1,circular=1","support","1","yes","CPU"
"CPU","FLASH_ATTN_EXT","hsk=40,hsv=40,nh=4,nr23=[1,1],kv=113,nb=1,mask=1,sinks=1,max_bias=0.000000,logit_softcap=0.000000,prec=f32,type_KV=f32,permute=[0,1,2,3]","support","1","yes","CPU"
"CPU","FLASH_ATTN_EXT","hsk=40,hsv=40,nh=4,nr23=[1,1],kv=113,nb=1,mask=1,sinks=1,max_bias=0.000000,logit_softcap=0.000000,prec=f32,type_KV=f16,permute=[0,1,2,3]","support","1","yes","CPU"
"CPU","FLASH_ATTN_EXT","hsk=40,hsv=40,nh=4,nr23=[1,1],kv=113,nb=1,mask=1,sinks=1,max_bias=0.000000,logit_softcap=0.000000,prec=f32,type_KV=bf16,permute=[0,1,2,3]","support","1","yes","CPU"

Can't render this file because it is too large.

524
docs/ops/CUDA.csv
View File

@ -4964,6 +4964,7 @@
"CUDA0","CONV_TRANSPOSE_1D","ne_input=[2,1,1,1],ne_kernel=[3,1,1,1],s0=1,p0=0,d0=1","support","1","yes","CUDA"
"CUDA0","CONV_TRANSPOSE_2D","ne_input=[3,2,3,1],ne_kernel=[2,2,1,3],stride=1","support","1","yes","CUDA"
"CUDA0","CONV_TRANSPOSE_2D","ne_input=[10,10,9,1],ne_kernel=[3,3,1,9],stride=2","support","1","yes","CUDA"
"CUDA0","CONV_TRANSPOSE_2D","ne_input=[129,63,35,1],ne_kernel=[3,3,48,35],stride=1","support","1","yes","CUDA"
"CUDA0","COUNT_EQUAL","type=f32,ne=[4,500,1,1]","support","1","yes","CUDA"
"CUDA0","COUNT_EQUAL","type=f32,ne=[4,5000,1,1]","support","1","yes","CUDA"
"CUDA0","ARGMAX","type=f32,ne=[32,1,1,1]","support","1","yes","CUDA"
@ -5419,17 +5420,45 @@
"CUDA0","CPY","type_src=f16,type_dst=f16,ne=[256,4,1,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","1","yes","CUDA"
"CUDA0","CPY","type_src=f32,type_dst=f32,ne=[256,4,1,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","1","yes","CUDA"
"CUDA0","CPY","type_src=bf16,type_dst=bf16,ne=[256,4,1,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","1","yes","CUDA"
"CUDA0","CPY","type_src=i32,type_dst=i32,ne=[256,4,1,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","1","yes","CUDA"
"CUDA0","CPY","type_src=i32,type_dst=i32,ne=[256,1,4,1],permute_src=[1,2,0,3],permute_dst=[0,0,0,0],_src_transpose=0","support","1","yes","CUDA"
"CUDA0","CPY","type_src=f32,type_dst=f32,ne=[256,1,4,1],permute_src=[1,2,0,3],permute_dst=[0,0,0,0],_src_transpose=0","support","1","yes","CUDA"
"CUDA0","CONT","type=f32,ne=[10,10,10,1]","support","1","yes","CUDA"
"CUDA0","CONT","type=f32,ne=[2,1,1,1]","support","1","yes","CUDA"
"CUDA0","CONT","type=f32,ne=[2,1,3,5]","support","1","yes","CUDA"
"CUDA0","CONT","type=f32,ne=[2,3,5,7]","support","1","yes","CUDA"
"CUDA0","CONT","type=f16,ne=[2,1,1,1]","support","1","yes","CUDA"
"CUDA0","CONT","type=f16,ne=[2,1,3,5]","support","1","yes","CUDA"
"CUDA0","CONT","type=f16,ne=[2,3,5,7]","support","1","yes","CUDA"
"CUDA0","CONT","type=bf16,ne=[2,1,1,1]","support","1","yes","CUDA"
"CUDA0","CONT","type=bf16,ne=[2,1,3,5]","support","1","yes","CUDA"
"CUDA0","CONT","type=bf16,ne=[2,3,5,7]","support","1","yes","CUDA"
"CUDA0","CONT","type=f32,ne=[2,1,1,1],use_view_slice=1","support","1","yes","CUDA"
"CUDA0","CONT","type=f32,ne=[2,1,3,5],use_view_slice=1","support","1","yes","CUDA"
"CUDA0","CONT","type=f32,ne=[2,3,5,7],use_view_slice=1","support","1","yes","CUDA"
"CUDA0","CONT","type=f32,ne=[1,4,4,1],use_view_slice=1","support","1","yes","CUDA"
"CUDA0","CONT","type=f32,ne=[1,8,17,1],use_view_slice=1","support","1","yes","CUDA"
"CUDA0","CONT","type=f32,ne=[10,10,10,1],use_view_slice=1","support","1","yes","CUDA"
"CUDA0","CONT","type=f32,ne=[2,1,1,1],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=f32,ne=[2,1,3,5],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=f32,ne=[2,3,5,7],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=f32,ne=[1,4,4,1],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=f32,ne=[1,8,17,1],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=f32,ne=[10,10,10,1],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=i32,ne=[2,1,1,1],use_view_slice=1","support","1","yes","CUDA"
"CUDA0","CONT","type=i32,ne=[2,1,3,5],use_view_slice=1","support","1","yes","CUDA"
"CUDA0","CONT","type=i32,ne=[2,3,5,7],use_view_slice=1","support","1","yes","CUDA"
"CUDA0","CONT","type=i32,ne=[1,4,4,1],use_view_slice=1","support","1","yes","CUDA"
"CUDA0","CONT","type=i32,ne=[1,8,17,1],use_view_slice=1","support","1","yes","CUDA"
"CUDA0","CONT","type=i32,ne=[10,10,10,1],use_view_slice=1","support","1","yes","CUDA"
"CUDA0","CONT","type=i32,ne=[2,1,1,1],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=i32,ne=[2,1,3,5],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=i32,ne=[2,3,5,7],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=i32,ne=[1,4,4,1],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=i32,ne=[1,8,17,1],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=i32,ne=[10,10,10,1],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=f16,ne=[2,1,1,1],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=f16,ne=[2,1,3,5],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=f16,ne=[2,3,5,7],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=f16,ne=[1,4,4,1],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=f16,ne=[1,8,17,1],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=f16,ne=[10,10,10,1],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=bf16,ne=[2,1,1,1],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=bf16,ne=[2,1,3,5],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=bf16,ne=[2,3,5,7],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=bf16,ne=[1,4,4,1],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=bf16,ne=[1,8,17,1],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","CONT","type=bf16,ne=[10,10,10,1],use_view_slice=0","support","1","yes","CUDA"
"CUDA0","ADD","type=f16,ne=[1,1,8,1],nr=[1,1,1,1],nf=1","support","1","yes","CUDA"
"CUDA0","SUB","type=f16,ne=[1,1,8,1],nr=[1,1,1,1],nf=1","support","1","yes","CUDA"
"CUDA0","MUL","type=f16,ne=[1,1,8,1],nr=[1,1,1,1],nf=1","support","1","yes","CUDA"
@ -5655,6 +5684,7 @@
"CUDA0","MUL","type=f32,ne=[64,262144,1,1],nr=[1,1,1,1],nf=1","support","1","yes","CUDA"
"CUDA0","DIV","type=f32,ne=[64,262144,1,1],nr=[1,1,1,1],nf=1","support","1","yes","CUDA"
"CUDA0","ADD1","type=f32,ne=[10,5,4,3]","support","1","yes","CUDA"
"CUDA0","ADD1","type=f32,ne=[1024,1024,1,1]","support","1","yes","CUDA"
"CUDA0","SCALE","type=f32,ne=[10,10,10,10],scale=2.000000,bias=0.000000,inplace=0","support","1","yes","CUDA"
"CUDA0","SCALE","type=f32,ne=[10,10,10,10],scale=2.000000,bias=1.000000,inplace=0","support","1","yes","CUDA"
"CUDA0","SCALE","type=f32,ne=[10,10,10,10],scale=2.000000,bias=1.000000,inplace=1","support","1","yes","CUDA"
@ -8644,9 +8674,13 @@
"CUDA0","CLAMP","type=f16,ne=[7,1,5,3],min=-0.500000,max=0.500000","support","1","yes","CUDA"
"CUDA0","LEAKY_RELU","type=f16,ne_a=[7,1,5,3],negative_slope=0.100000","support","1","yes","CUDA"
"CUDA0","FLOOR","type=f16,ne=[7,1,5,3]","support","1","yes","CUDA"
"CUDA0","FLOOR","type=f16,ne=[1024,1024,1,1]","support","1","yes","CUDA"
"CUDA0","CEIL","type=f16,ne=[7,1,5,3]","support","1","yes","CUDA"
"CUDA0","CEIL","type=f16,ne=[1024,1024,1,1]","support","1","yes","CUDA"
"CUDA0","ROUND","type=f16,ne=[7,1,5,3]","support","1","yes","CUDA"
"CUDA0","ROUND","type=f16,ne=[1024,1024,1,1]","support","1","yes","CUDA"
"CUDA0","TRUNC","type=f16,ne=[7,1,5,3]","support","1","yes","CUDA"
"CUDA0","TRUNC","type=f16,ne=[1024,1024,1,1]","support","1","yes","CUDA"
"CUDA0","SQR","type=f32,ne=[10,5,4,3]","support","1","yes","CUDA"
"CUDA0","SQRT","type=f32,ne=[10,3,3,2]","support","1","yes","CUDA"
"CUDA0","LOG","type=f32,ne=[10,5,4,3]","support","1","yes","CUDA"
@ -8666,9 +8700,13 @@
"CUDA0","CLAMP","type=f32,ne=[7,1,5,3],min=-0.500000,max=0.500000","support","1","yes","CUDA"
"CUDA0","LEAKY_RELU","type=f32,ne_a=[7,1,5,3],negative_slope=0.100000","support","1","yes","CUDA"
"CUDA0","FLOOR","type=f32,ne=[7,1,5,3]","support","1","yes","CUDA"
"CUDA0","FLOOR","type=f32,ne=[1024,1024,1,1]","support","1","yes","CUDA"
"CUDA0","CEIL","type=f32,ne=[7,1,5,3]","support","1","yes","CUDA"
"CUDA0","CEIL","type=f32,ne=[1024,1024,1,1]","support","1","yes","CUDA"
"CUDA0","ROUND","type=f32,ne=[7,1,5,3]","support","1","yes","CUDA"
"CUDA0","ROUND","type=f32,ne=[1024,1024,1,1]","support","1","yes","CUDA"
"CUDA0","TRUNC","type=f32,ne=[7,1,5,3]","support","1","yes","CUDA"
"CUDA0","TRUNC","type=f32,ne=[1024,1024,1,1]","support","1","yes","CUDA"
"CUDA0","DIAG_MASK_INF","type=f32,ne=[10,10,1,1],n_past=5","support","1","yes","CUDA"
"CUDA0","DIAG_MASK_INF","type=f32,ne=[10,10,3,1],n_past=5","support","1","yes","CUDA"
"CUDA0","DIAG_MASK_INF","type=f32,ne=[10,10,3,2],n_past=5","support","1","yes","CUDA"
@ -9411,18 +9449,405 @@
"CUDA0","CONCAT","type=i32,ne_a=[11,12,13,14],ne_b_d=7,dim=2,v=3","support","0","no","CUDA"
"CUDA0","CONCAT","type=f32,ne_a=[11,12,13,14],ne_b_d=7,dim=3,v=3","support","1","yes","CUDA"
"CUDA0","CONCAT","type=i32,ne_a=[11,12,13,14],ne_b_d=7,dim=3,v=3","support","0","no","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[3,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[4,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[7,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[8,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[15,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[16,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[31,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[32,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[63,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[64,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[127,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[128,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[255,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[256,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[511,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[512,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[1023,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[1024,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[2047,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[2048,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[4095,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[4096,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[8191,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[8192,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[16383,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[16384,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[32767,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[32768,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[65535,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[65536,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[131071,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[131072,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[262143,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[262144,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[524287,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[524288,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[1048575,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[1048576,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[16,10,10,10],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[60,10,10,10],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[1024,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[16384,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[1023,2,1,3],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[1024,2,1,3],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[1025,2,1,3],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[2047,2,1,3],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[2048,2,1,3],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[2049,2,1,3],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[2,8,8192,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[8,1,1,1],order=1","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[3,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[4,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[7,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[8,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[15,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[16,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[31,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[32,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[63,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[64,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[127,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[128,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[255,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[256,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[511,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[512,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[1023,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[1024,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[2047,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[2048,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[4095,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[4096,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[8191,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[8192,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[16383,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[16384,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[32767,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[32768,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[65535,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[65536,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[131071,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[131072,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[262143,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[262144,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[524287,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[524288,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[1048575,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[1048576,1,1,1],order=0","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[16,10,10,10],order=1","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[60,10,10,10],order=1","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[1024,1,1,1],order=1","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[16384,1,1,1],order=1","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[1023,2,1,3],order=1","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[1024,2,1,3],order=1","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[1025,2,1,3],order=1","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[2047,2,1,3],order=1","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[2048,2,1,3],order=1","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[2049,2,1,3],order=1","support","1","yes","CUDA"
"CUDA0","ARGSORT","type=f32,ne=[2,8,8192,1],order=1","support","1","yes","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[12,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[13,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[13,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[15,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[15,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[15,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[19,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[19,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[19,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[19,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[27,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[27,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[27,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[27,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[27,1,2,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[43,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[43,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[43,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[43,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[43,1,2,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[64,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[75,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[64,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[75,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[64,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[75,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[64,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[75,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[64,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[75,1,2,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[128,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[139,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[128,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[139,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[128,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[139,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[128,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[139,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[128,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[139,1,2,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[128,1,1,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[139,1,2,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[256,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[267,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[256,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[267,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[256,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[267,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[256,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[267,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[256,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[267,1,2,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[256,1,1,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[267,1,2,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[512,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[523,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[512,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[523,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[512,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[523,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[512,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[523,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[512,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[523,1,2,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[512,1,1,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[523,1,2,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[512,1,1,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[523,1,2,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1024,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1035,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1024,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1035,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1024,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1035,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1024,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1035,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1024,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1035,1,2,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1024,1,1,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1035,1,2,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1024,1,1,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1035,1,2,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1024,1,1,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1035,1,2,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2048,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2059,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2048,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2059,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2048,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2059,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2048,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2059,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2048,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2059,1,2,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2048,1,1,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2059,1,2,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2048,1,1,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2059,1,2,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2048,1,1,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2059,1,2,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4096,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4107,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4096,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4107,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4096,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4107,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4096,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4107,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4096,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4107,1,2,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4096,1,1,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4107,1,2,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4096,1,1,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4107,1,2,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4096,1,1,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[4107,1,2,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8192,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8203,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8192,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8203,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8192,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8203,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8192,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8203,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8192,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8203,1,2,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8192,1,1,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8203,1,2,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8192,1,1,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8203,1,2,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8192,1,1,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[8203,1,2,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16384,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16395,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16384,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16395,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16384,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16395,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16384,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16395,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16384,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16395,1,2,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16384,1,1,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16395,1,2,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16384,1,1,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16395,1,2,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16384,1,1,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16395,1,2,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16384,1,1,1],k=9999,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16395,1,2,1],k=9999,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32768,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32779,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32768,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32779,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32768,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32779,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32768,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32779,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32768,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32779,1,2,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32768,1,1,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32779,1,2,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32768,1,1,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32779,1,2,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32768,1,1,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32779,1,2,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32768,1,1,1],k=9999,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[32779,1,2,1],k=9999,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65536,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65547,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65536,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65547,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65536,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65547,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65536,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65547,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65536,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65547,1,2,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65536,1,1,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65547,1,2,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65536,1,1,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65547,1,2,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65536,1,1,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65547,1,2,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65536,1,1,1],k=9999,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[65547,1,2,1],k=9999,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131072,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131083,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131072,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131083,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131072,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131083,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131072,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131083,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131072,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131083,1,2,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131072,1,1,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131083,1,2,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131072,1,1,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131083,1,2,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131072,1,1,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131083,1,2,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131072,1,1,1],k=9999,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[131083,1,2,1],k=9999,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262144,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262155,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262144,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262155,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262144,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262155,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262144,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262155,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262144,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262155,1,2,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262144,1,1,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262155,1,2,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262144,1,1,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262155,1,2,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262144,1,1,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262155,1,2,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262144,1,1,1],k=9999,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[262155,1,2,1],k=9999,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524288,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524299,1,2,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524288,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524299,1,2,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524288,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524299,1,2,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524288,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524299,1,2,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524288,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524299,1,2,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524288,1,1,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524299,1,2,1],k=100,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524288,1,1,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524299,1,2,1],k=500,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524288,1,1,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524299,1,2,1],k=1023,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524288,1,1,1],k=9999,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[524299,1,2,1],k=9999,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16,10,10,10],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[60,10,10,10],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1023,2,1,3],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1024,2,1,3],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1025,2,1,3],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16384,1,1,1],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2047,2,1,3],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2048,2,1,3],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2049,2,1,3],k=1,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16,10,10,10],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[60,10,10,10],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1023,2,1,3],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1024,2,1,3],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1025,2,1,3],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16384,1,1,1],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2047,2,1,3],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2048,2,1,3],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2049,2,1,3],k=2,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16,10,10,10],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[60,10,10,10],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1023,2,1,3],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1024,2,1,3],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1025,2,1,3],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16384,1,1,1],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2047,2,1,3],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2048,2,1,3],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2049,2,1,3],k=3,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16,10,10,10],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[60,10,10,10],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1023,2,1,3],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1024,2,1,3],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1025,2,1,3],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16384,1,1,1],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2047,2,1,3],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2048,2,1,3],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2049,2,1,3],k=7,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16,10,10,10],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[60,10,10,10],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1023,2,1,3],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1024,2,1,3],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[1025,2,1,3],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[16384,1,1,1],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2047,2,1,3],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2048,2,1,3],k=15,ties=0","support","0","no","CUDA"
"CUDA0","TOP_K","type=f32,ne=[2049,2,1,3],k=15,ties=0","support","0","no","CUDA"
"CUDA0","UPSCALE","type=f32,ne=[512,512,3,2],scale_factor=2,mode=nearest,transpose=0","support","1","yes","CUDA"
"CUDA0","UPSCALE","type=f32,ne=[512,512,3,2],scale_factor=2,mode=nearest,transpose=1","support","1","yes","CUDA"
"CUDA0","UPSCALE","type=f32,ne=[2,5,7,11],ne_tgt=[5,7,11,13],mode=nearest,flags=none","support","1","yes","CUDA"
@ -9435,6 +9860,10 @@
"CUDA0","UPSCALE","type=f32,ne=[512,512,3,2],scale_factor=2,mode=bicubic,transpose=1","support","1","yes","CUDA"
"CUDA0","UPSCALE","type=f32,ne=[2,5,7,11],ne_tgt=[5,7,11,13],mode=bicubic,flags=none","support","1","yes","CUDA"
"CUDA0","UPSCALE","type=f32,ne=[5,7,11,13],ne_tgt=[2,5,7,11],mode=bicubic,flags=none","support","1","yes","CUDA"
"CUDA0","UPSCALE","type=f32,ne=[512,512,3,2],scale_factor=2,mode=513,transpose=0","support","1","yes","CUDA"
"CUDA0","UPSCALE","type=f32,ne=[512,512,3,2],scale_factor=2,mode=513,transpose=1","support","1","yes","CUDA"
"CUDA0","UPSCALE","type=f32,ne=[2,5,7,11],ne_tgt=[5,7,11,13],mode=bilinear,flags=none","support","1","yes","CUDA"
"CUDA0","UPSCALE","type=f32,ne=[5,7,11,13],ne_tgt=[2,5,7,11],mode=bilinear,flags=none","support","1","yes","CUDA"
"CUDA0","UPSCALE","type=f32,ne=[2,5,7,11],ne_tgt=[5,7,11,13],mode=bilinear,flags=align_corners","support","1","yes","CUDA"
"CUDA0","UPSCALE","type=f32,ne=[1,4,3,2],ne_tgt=[2,8,3,2],mode=bilinear,flags=align_corners","support","1","yes","CUDA"
"CUDA0","UPSCALE","type=f32,ne=[4,1,3,2],ne_tgt=[1,1,3,2],mode=bilinear,flags=align_corners","support","1","yes","CUDA"
@ -9463,34 +9892,59 @@
"CUDA0","GROUP_NORM","type=f32,ne=[64,64,320,1],num_groups=32,eps=0.000001","support","1","yes","CUDA"
"CUDA0","GROUP_NORM","type=f32,ne=[9,9,1280,1],num_groups=32,eps=0.000001","support","1","yes","CUDA"
"CUDA0","ACC","type=f32,ne_a=[256,17,1,1],ne_b=[256,16,1,1]","support","1","yes","CUDA"
"CUDA0","PAD","type=f32,ne_a=[512,512,1,1],pad_0=1,pad_1=1","support","1","yes","CUDA"
"CUDA0","PAD","type=f32,ne_a=[512,512,3,1],lp0=1,rp0=1,lp1=1,rp1=1,lp2=1,rp2=1,lp3=1,rp3=1,v=0","support","1","yes","CUDA"
"CUDA0","PAD","type=f32,ne_a=[512,512,1,1],pad_0=1,pad_1=1,circular=0","support","1","yes","CUDA"
"CUDA0","PAD","type=f32,ne_a=[33,17,2,1],pad_0=4,pad_1=3,circular=1","support","1","yes","CUDA"
"CUDA0","PAD","type=f32,ne_a=[512,512,3,1],lp0=1,rp0=1,lp1=1,rp1=1,lp2=1,rp2=1,lp3=1,rp3=1,v=0,circular=0","support","1","yes","CUDA"
"CUDA0","PAD_REFLECT_1D","type=f32,ne_a=[512,34,2,1],pad_0=10,pad_1=9","support","1","yes","CUDA"
"CUDA0","PAD_REFLECT_1D","type=f32,ne_a=[3000,384,4,1],pad_0=10,pad_1=9","support","1","yes","CUDA"
"CUDA0","ROLL","shift0=3,shift1=-2,shift3=1,shift4=-1","support","1","yes","CUDA"
"CUDA0","ARANGE","type=f32,start=0.000000,stop=10.000000,step=1.000000","support","1","yes","CUDA"
"CUDA0","ARANGE","type=f32,start=0.000000,stop=1048576.000000,step=1.000000","support","1","yes","CUDA"
"CUDA0","TIMESTEP_EMBEDDING","type=f32,ne_a=[2,1,1,1],dim=320,max_period=10000","support","1","yes","CUDA"
"CUDA0","LEAKY_RELU","type=f32,ne_a=[10,5,4,3],negative_slope=0.100000","support","1","yes","CUDA"
"CUDA0","CUMSUM","type=f32,ne=[10,5,4,3]","support","0","no","CUDA"
"CUDA0","CUMSUM","type=f32,ne=[10,5,4,3]","support","1","yes","CUDA"
"CUDA0","CUMSUM","type=f32,ne=[127,5,4,3]","support","1","yes","CUDA"
"CUDA0","CUMSUM","type=f32,ne=[128,5,4,3]","support","1","yes","CUDA"
"CUDA0","CUMSUM","type=f32,ne=[128,128,4,4]","support","1","yes","CUDA"
"CUDA0","CUMSUM","type=f32,ne=[255,5,4,3]","support","1","yes","CUDA"
"CUDA0","CUMSUM","type=f32,ne=[256,5,4,3]","support","1","yes","CUDA"
"CUDA0","CUMSUM","type=f32,ne=[511,5,4,3]","support","1","yes","CUDA"
"CUDA0","CUMSUM","type=f32,ne=[512,5,4,3]","support","1","yes","CUDA"
"CUDA0","CUMSUM","type=f32,ne=[1023,5,4,3]","support","1","yes","CUDA"
"CUDA0","CUMSUM","type=f32,ne=[1024,5,4,3]","support","1","yes","CUDA"
"CUDA0","CUMSUM","type=f32,ne=[2047,5,4,3]","support","1","yes","CUDA"
"CUDA0","CUMSUM","type=f32,ne=[2048,5,4,3]","support","1","yes","CUDA"
"CUDA0","CUMSUM","type=f32,ne=[242004,1,1,1]","support","1","yes","CUDA"
"CUDA0","CUMSUM","type=f32,ne=[375960,1,1,1]","support","1","yes","CUDA"
"CUDA0","XIELU","type=f32,ne=[10,5,4,3]","support","0","no","CUDA"
"CUDA0","TRI","type=f32,ne=[10,10,4,3],tri_type=3","support","0","no","CUDA"
"CUDA0","TRI","type=f32,ne=[10,10,4,3],tri_type=2","support","0","no","CUDA"
"CUDA0","TRI","type=f32,ne=[10,10,4,3],tri_type=1","support","0","no","CUDA"
"CUDA0","TRI","type=f32,ne=[10,10,4,3],tri_type=0","support","0","no","CUDA"
"CUDA0","FILL","type=f32,ne=[10,10,4,3],c=0.000000","support","0","no","CUDA"
"CUDA0","FILL","type=f32,ne=[303,207,11,3],c=2.000000","support","0","no","CUDA"
"CUDA0","FILL","type=f32,ne=[800,600,4,4],c=-152.000000","support","0","no","CUDA"
"CUDA0","SOLVE_TRI","type=f32,ne_lhs=[10,10,4,3],ne_rhs=[3,10,4,3]","support","0","no","CUDA"
"CUDA0","SOLVE_TRI","type=f32,ne_lhs=[11,11,1,1],ne_rhs=[5,11,1,1]","support","0","no","CUDA"
"CUDA0","SOLVE_TRI","type=f32,ne_lhs=[17,17,2,4],ne_rhs=[9,17,2,4]","support","0","no","CUDA"
"CUDA0","SOLVE_TRI","type=f32,ne_lhs=[30,30,7,1],ne_rhs=[8,30,7,1]","support","0","no","CUDA"
"CUDA0","SOLVE_TRI","type=f32,ne_lhs=[42,42,5,2],ne_rhs=[10,42,5,2]","support","0","no","CUDA"
"CUDA0","SOLVE_TRI","type=f32,ne_lhs=[64,64,2,2],ne_rhs=[10,64,2,2]","support","0","no","CUDA"
"CUDA0","TRI","type=f32,ne=[10,10,4,3],tri_type=3","support","1","yes","CUDA"
"CUDA0","TRI","type=f32,ne=[10,10,4,3],tri_type=2","support","1","yes","CUDA"
"CUDA0","TRI","type=f32,ne=[10,10,4,3],tri_type=1","support","1","yes","CUDA"
"CUDA0","TRI","type=f32,ne=[10,10,4,3],tri_type=0","support","1","yes","CUDA"
"CUDA0","FILL","type=f32,ne=[10,10,4,3],c=0.000000","support","1","yes","CUDA"
"CUDA0","FILL","type=f32,ne=[303,207,11,3],c=2.000000","support","1","yes","CUDA"
"CUDA0","FILL","type=f32,ne=[800,600,4,4],c=-152.000000","support","1","yes","CUDA"
"CUDA0","FILL","type=f32,ne=[2048,512,2,2],c=3.500000","support","1","yes","CUDA"
"CUDA0","DIAG","type=f32,ne=[10,1,4,3]","support","1","yes","CUDA"
"CUDA0","DIAG","type=f32,ne=[79,1,19,13]","support","1","yes","CUDA"
"CUDA0","DIAG","type=f32,ne=[256,1,8,16]","support","1","yes","CUDA"
"CUDA0","SOLVE_TRI","type=f32,ne_lhs=[10,10,4,3],ne_rhs=[3,10,4,3]","support","1","yes","CUDA"
"CUDA0","SOLVE_TRI","type=f32,ne_lhs=[11,11,1,1],ne_rhs=[5,11,1,1]","support","1","yes","CUDA"
"CUDA0","SOLVE_TRI","type=f32,ne_lhs=[17,17,2,4],ne_rhs=[9,17,2,4]","support","1","yes","CUDA"
"CUDA0","SOLVE_TRI","type=f32,ne_lhs=[30,30,7,1],ne_rhs=[8,30,7,1]","support","1","yes","CUDA"
"CUDA0","SOLVE_TRI","type=f32,ne_lhs=[42,42,5,2],ne_rhs=[10,42,5,2]","support","1","yes","CUDA"
"CUDA0","SOLVE_TRI","type=f32,ne_lhs=[64,64,2,2],ne_rhs=[10,64,2,2]","support","1","yes","CUDA"
"CUDA0","SOLVE_TRI","type=f32,ne_lhs=[100,100,4,4],ne_rhs=[41,100,4,4]","support","0","no","CUDA"
"CUDA0","PAD","type=f32,ne_a=[512,512,1,1],lp0=0,rp0=1,lp1=0,rp1=1,lp2=0,rp2=0,lp3=0,rp3=0,v=0","support","1","yes","CUDA"
"CUDA0","PAD","type=f32,ne_a=[11,22,33,44],lp0=1,rp0=2,lp1=3,rp1=4,lp2=5,rp2=6,lp3=7,rp3=8,v=0","support","1","yes","CUDA"
"CUDA0","PAD","type=f32,ne_a=[512,512,1,1],lp0=0,rp0=1,lp1=0,rp1=1,lp2=0,rp2=0,lp3=0,rp3=0,v=1","support","0","no","CUDA"
"CUDA0","PAD","type=f32,ne_a=[11,22,33,44],lp0=1,rp0=2,lp1=3,rp1=4,lp2=5,rp2=6,lp3=7,rp3=8,v=1","support","0","no","CUDA"
"CUDA0","SOLVE_TRI","type=f32,ne_lhs=[128,128,4,4],ne_rhs=[31,128,4,4]","support","0","no","CUDA"
"CUDA0","SOLVE_TRI","type=f32,ne_lhs=[64,64,4,4],ne_rhs=[300,64,4,4]","support","0","no","CUDA"
"CUDA0","PAD","type=f32,ne_a=[512,512,1,1],lp0=0,rp0=1,lp1=0,rp1=1,lp2=0,rp2=0,lp3=0,rp3=0,v=0,circular=0","support","1","yes","CUDA"
"CUDA0","PAD","type=f32,ne_a=[11,22,33,44],lp0=1,rp0=2,lp1=3,rp1=4,lp2=5,rp2=6,lp3=7,rp3=8,v=0,circular=0","support","1","yes","CUDA"
"CUDA0","PAD","type=f32,ne_a=[512,512,1,1],lp0=0,rp0=1,lp1=0,rp1=1,lp2=0,rp2=0,lp3=0,rp3=0,v=0,circular=1","support","1","yes","CUDA"
"CUDA0","PAD","type=f32,ne_a=[11,22,33,44],lp0=1,rp0=2,lp1=3,rp1=4,lp2=5,rp2=6,lp3=7,rp3=8,v=0,circular=1","support","1","yes","CUDA"
"CUDA0","PAD","type=f32,ne_a=[512,512,1,1],lp0=0,rp0=1,lp1=0,rp1=1,lp2=0,rp2=0,lp3=0,rp3=0,v=1,circular=0","support","0","no","CUDA"
"CUDA0","PAD","type=f32,ne_a=[11,22,33,44],lp0=1,rp0=2,lp1=3,rp1=4,lp2=5,rp2=6,lp3=7,rp3=8,v=1,circular=0","support","0","no","CUDA"
"CUDA0","PAD","type=f32,ne_a=[512,512,1,1],lp0=0,rp0=1,lp1=0,rp1=1,lp2=0,rp2=0,lp3=0,rp3=0,v=1,circular=1","support","0","no","CUDA"
"CUDA0","PAD","type=f32,ne_a=[11,22,33,44],lp0=1,rp0=2,lp1=3,rp1=4,lp2=5,rp2=6,lp3=7,rp3=8,v=1,circular=1","support","0","no","CUDA"
"CUDA0","FLASH_ATTN_EXT","hsk=40,hsv=40,nh=4,nr23=[1,1],kv=113,nb=1,mask=1,sinks=1,max_bias=0.000000,logit_softcap=0.000000,prec=f32,type_KV=f32,permute=[0,1,2,3]","support","1","yes","CUDA"
"CUDA0","FLASH_ATTN_EXT","hsk=40,hsv=40,nh=4,nr23=[1,1],kv=113,nb=1,mask=1,sinks=1,max_bias=0.000000,logit_softcap=0.000000,prec=f32,type_KV=f16,permute=[0,1,2,3]","support","1","yes","CUDA"
"CUDA0","FLASH_ATTN_EXT","hsk=40,hsv=40,nh=4,nr23=[1,1],kv=113,nb=1,mask=1,sinks=1,max_bias=0.000000,logit_softcap=0.000000,prec=f32,type_KV=bf16,permute=[0,1,2,3]","support","0","no","CUDA"

Can't render this file because it is too large.

22617
docs/ops/Metal.csv
View File

File diff suppressed because it is too large Load Diff

19640
docs/ops/OpenCL.csv
View File

File diff suppressed because it is too large Load Diff

1158
docs/ops/SYCL.csv
View File

File diff suppressed because it is too large Load Diff

518
docs/ops/Vulkan.csv
View File

@ -5005,8 +5005,8 @@
"Vulkan0","DUP","type=f16,ne=[10,10,5,1],permute=[0,2,1,3]","support","1","yes","Vulkan"
"Vulkan0","DUP","type=f32,ne=[10,10,5,1],permute=[1,0,2,3]","support","1","yes","Vulkan"
"Vulkan0","DUP","type=f16,ne=[10,10,5,1],permute=[1,0,2,3]","support","1","yes","Vulkan"
"Vulkan0","DUP","type=i16,ne=[10,8,3,1],permute=[0,2,1,3]","support","0","no","Vulkan"
"Vulkan0","DUP","type=i16,ne=[10,8,3,1],permute=[1,2,0,3]","support","0","no","Vulkan"
"Vulkan0","DUP","type=i16,ne=[10,8,3,1],permute=[0,2,1,3]","support","1","yes","Vulkan"
"Vulkan0","DUP","type=i16,ne=[10,8,3,1],permute=[1,2,0,3]","support","1","yes","Vulkan"
"Vulkan0","SET","type_src=f32,type_dst=f32,ne=[6,5,4,3],dim=1","support","0","no","Vulkan"
"Vulkan0","SET","type_src=f32,type_dst=f32,ne=[6,5,4,3],dim=2","support","0","no","Vulkan"
"Vulkan0","SET","type_src=f32,type_dst=f32,ne=[6,5,4,3],dim=3","support","0","no","Vulkan"
@ -5032,14 +5032,14 @@
"Vulkan0","CPY","type_src=f16,type_dst=f16,ne=[3,2,3,4],permute_src=[0,2,1,3],permute_dst=[0,0,0,0],_src_transpose=0","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=f16,type_dst=f16,ne=[3,2,3,4],permute_src=[0,3,1,2],permute_dst=[0,2,1,3],_src_transpose=0","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[1,2,3,4],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=0","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[1,2,3,4],permute_src=[0,2,1,3],permute_dst=[0,0,0,0],_src_transpose=0","support","0","no","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[1,2,3,4],permute_src=[0,3,1,2],permute_dst=[0,2,1,3],_src_transpose=0","support","0","no","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[1,2,3,4],permute_src=[0,2,1,3],permute_dst=[0,0,0,0],_src_transpose=0","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[1,2,3,4],permute_src=[0,3,1,2],permute_dst=[0,2,1,3],_src_transpose=0","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[2,2,3,4],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=0","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[2,2,3,4],permute_src=[0,2,1,3],permute_dst=[0,0,0,0],_src_transpose=0","support","0","no","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[2,2,3,4],permute_src=[0,3,1,2],permute_dst=[0,2,1,3],_src_transpose=0","support","0","no","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[2,2,3,4],permute_src=[0,2,1,3],permute_dst=[0,0,0,0],_src_transpose=0","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[2,2,3,4],permute_src=[0,3,1,2],permute_dst=[0,2,1,3],_src_transpose=0","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[3,2,3,4],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=0","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[3,2,3,4],permute_src=[0,2,1,3],permute_dst=[0,0,0,0],_src_transpose=0","support","0","no","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[3,2,3,4],permute_src=[0,3,1,2],permute_dst=[0,2,1,3],_src_transpose=0","support","0","no","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[3,2,3,4],permute_src=[0,2,1,3],permute_dst=[0,0,0,0],_src_transpose=0","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[3,2,3,4],permute_src=[0,3,1,2],permute_dst=[0,2,1,3],_src_transpose=0","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=q4_0,type_dst=q4_0,ne=[32,2,3,4],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=0","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=q4_0,type_dst=q4_0,ne=[32,2,3,4],permute_src=[0,2,1,3],permute_dst=[0,0,0,0],_src_transpose=0","support","0","no","Vulkan"
"Vulkan0","CPY","type_src=q4_0,type_dst=q4_0,ne=[32,2,3,4],permute_src=[0,3,1,2],permute_dst=[0,2,1,3],_src_transpose=0","support","0","no","Vulkan"
@ -5271,7 +5271,7 @@
"Vulkan0","CPY","type_src=bf16,type_dst=f16,ne=[256,4,4,4],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=0","support","0","no","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=f16,ne=[256,2,3,4],permute_src=[0,2,1,3],permute_dst=[0,0,0,0],_src_transpose=0","support","0","no","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[256,4,4,4],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=0","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[256,2,3,4],permute_src=[0,2,1,3],permute_dst=[0,0,0,0],_src_transpose=0","support","0","no","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[256,2,3,4],permute_src=[0,2,1,3],permute_dst=[0,0,0,0],_src_transpose=0","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=q4_0,ne=[256,4,4,4],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=0","support","0","no","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=q4_0,ne=[256,2,3,4],permute_src=[0,2,1,3],permute_dst=[0,0,0,0],_src_transpose=0","support","0","no","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=q4_1,ne=[256,4,4,4],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=0","support","0","no","Vulkan"
@ -5415,21 +5415,49 @@
"Vulkan0","CPY","type_src=f16,type_dst=f16,ne=[256,4,3,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=f32,type_dst=f32,ne=[256,4,3,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=f32,type_dst=f32,ne=[256,4,3,3],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[256,4,3,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","0","no","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[256,4,3,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=f16,type_dst=f16,ne=[256,4,1,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=f32,type_dst=f32,ne=[256,4,1,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[256,4,1,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","0","no","Vulkan"
"Vulkan0","CPY","type_src=bf16,type_dst=bf16,ne=[256,4,1,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=i32,type_dst=i32,ne=[256,4,1,1],permute_src=[0,0,0,0],permute_dst=[0,0,0,0],_src_transpose=1","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=i32,type_dst=i32,ne=[256,1,4,1],permute_src=[1,2,0,3],permute_dst=[0,0,0,0],_src_transpose=0","support","1","yes","Vulkan"
"Vulkan0","CPY","type_src=f32,type_dst=f32,ne=[256,1,4,1],permute_src=[1,2,0,3],permute_dst=[0,0,0,0],_src_transpose=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f32,ne=[10,10,10,1]","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f32,ne=[2,1,1,1]","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f32,ne=[2,1,3,5]","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f32,ne=[2,3,5,7]","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f16,ne=[2,1,1,1]","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f16,ne=[2,1,3,5]","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f16,ne=[2,3,5,7]","support","1","yes","Vulkan"
"Vulkan0","CONT","type=bf16,ne=[2,1,1,1]","support","1","yes","Vulkan"
"Vulkan0","CONT","type=bf16,ne=[2,1,3,5]","support","1","yes","Vulkan"
"Vulkan0","CONT","type=bf16,ne=[2,3,5,7]","support","0","no","Vulkan"
"Vulkan0","CONT","type=f32,ne=[2,1,1,1],use_view_slice=1","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f32,ne=[2,1,3,5],use_view_slice=1","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f32,ne=[2,3,5,7],use_view_slice=1","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f32,ne=[1,4,4,1],use_view_slice=1","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f32,ne=[1,8,17,1],use_view_slice=1","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f32,ne=[10,10,10,1],use_view_slice=1","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f32,ne=[2,1,1,1],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f32,ne=[2,1,3,5],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f32,ne=[2,3,5,7],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f32,ne=[1,4,4,1],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f32,ne=[1,8,17,1],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f32,ne=[10,10,10,1],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=i32,ne=[2,1,1,1],use_view_slice=1","support","1","yes","Vulkan"
"Vulkan0","CONT","type=i32,ne=[2,1,3,5],use_view_slice=1","support","1","yes","Vulkan"
"Vulkan0","CONT","type=i32,ne=[2,3,5,7],use_view_slice=1","support","1","yes","Vulkan"
"Vulkan0","CONT","type=i32,ne=[1,4,4,1],use_view_slice=1","support","1","yes","Vulkan"
"Vulkan0","CONT","type=i32,ne=[1,8,17,1],use_view_slice=1","support","1","yes","Vulkan"
"Vulkan0","CONT","type=i32,ne=[10,10,10,1],use_view_slice=1","support","1","yes","Vulkan"
"Vulkan0","CONT","type=i32,ne=[2,1,1,1],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=i32,ne=[2,1,3,5],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=i32,ne=[2,3,5,7],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=i32,ne=[1,4,4,1],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=i32,ne=[1,8,17,1],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=i32,ne=[10,10,10,1],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f16,ne=[2,1,1,1],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f16,ne=[2,1,3,5],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f16,ne=[2,3,5,7],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f16,ne=[1,4,4,1],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f16,ne=[1,8,17,1],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=f16,ne=[10,10,10,1],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=bf16,ne=[2,1,1,1],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=bf16,ne=[2,1,3,5],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=bf16,ne=[2,3,5,7],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=bf16,ne=[1,4,4,1],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=bf16,ne=[1,8,17,1],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","CONT","type=bf16,ne=[10,10,10,1],use_view_slice=0","support","1","yes","Vulkan"
"Vulkan0","ADD","type=f16,ne=[1,1,8,1],nr=[1,1,1,1],nf=1","support","1","yes","Vulkan"
"Vulkan0","SUB","type=f16,ne=[1,1,8,1],nr=[1,1,1,1],nf=1","support","1","yes","Vulkan"
"Vulkan0","MUL","type=f16,ne=[1,1,8,1],nr=[1,1,1,1],nf=1","support","1","yes","Vulkan"
@ -5655,6 +5683,7 @@
"Vulkan0","MUL","type=f32,ne=[64,262144,1,1],nr=[1,1,1,1],nf=1","support","1","yes","Vulkan"
"Vulkan0","DIV","type=f32,ne=[64,262144,1,1],nr=[1,1,1,1],nf=1","support","1","yes","Vulkan"
"Vulkan0","ADD1","type=f32,ne=[10,5,4,3]","support","1","yes","Vulkan"
"Vulkan0","ADD1","type=f32,ne=[1024,1024,1,1]","support","1","yes","Vulkan"
"Vulkan0","SCALE","type=f32,ne=[10,10,10,10],scale=2.000000,bias=0.000000,inplace=0","support","1","yes","Vulkan"
"Vulkan0","SCALE","type=f32,ne=[10,10,10,10],scale=2.000000,bias=1.000000,inplace=0","support","1","yes","Vulkan"
"Vulkan0","SCALE","type=f32,ne=[10,10,10,10],scale=2.000000,bias=1.000000,inplace=1","support","1","yes","Vulkan"
@ -8644,9 +8673,13 @@
"Vulkan0","CLAMP","type=f16,ne=[7,1,5,3],min=-0.500000,max=0.500000","support","0","no","Vulkan"
"Vulkan0","LEAKY_RELU","type=f16,ne_a=[7,1,5,3],negative_slope=0.100000","support","0","no","Vulkan"
"Vulkan0","FLOOR","type=f16,ne=[7,1,5,3]","support","1","yes","Vulkan"
"Vulkan0","FLOOR","type=f16,ne=[1024,1024,1,1]","support","1","yes","Vulkan"
"Vulkan0","CEIL","type=f16,ne=[7,1,5,3]","support","1","yes","Vulkan"
"Vulkan0","CEIL","type=f16,ne=[1024,1024,1,1]","support","1","yes","Vulkan"
"Vulkan0","ROUND","type=f16,ne=[7,1,5,3]","support","1","yes","Vulkan"
"Vulkan0","ROUND","type=f16,ne=[1024,1024,1,1]","support","1","yes","Vulkan"
"Vulkan0","TRUNC","type=f16,ne=[7,1,5,3]","support","1","yes","Vulkan"
"Vulkan0","TRUNC","type=f16,ne=[1024,1024,1,1]","support","1","yes","Vulkan"
"Vulkan0","SQR","type=f32,ne=[10,5,4,3]","support","1","yes","Vulkan"
"Vulkan0","SQRT","type=f32,ne=[10,3,3,2]","support","1","yes","Vulkan"
"Vulkan0","LOG","type=f32,ne=[10,5,4,3]","support","1","yes","Vulkan"
@ -8666,9 +8699,13 @@
"Vulkan0","CLAMP","type=f32,ne=[7,1,5,3],min=-0.500000,max=0.500000","support","1","yes","Vulkan"
"Vulkan0","LEAKY_RELU","type=f32,ne_a=[7,1,5,3],negative_slope=0.100000","support","1","yes","Vulkan"
"Vulkan0","FLOOR","type=f32,ne=[7,1,5,3]","support","1","yes","Vulkan"
"Vulkan0","FLOOR","type=f32,ne=[1024,1024,1,1]","support","1","yes","Vulkan"
"Vulkan0","CEIL","type=f32,ne=[7,1,5,3]","support","1","yes","Vulkan"
"Vulkan0","CEIL","type=f32,ne=[1024,1024,1,1]","support","1","yes","Vulkan"
"Vulkan0","ROUND","type=f32,ne=[7,1,5,3]","support","1","yes","Vulkan"
"Vulkan0","ROUND","type=f32,ne=[1024,1024,1,1]","support","1","yes","Vulkan"
"Vulkan0","TRUNC","type=f32,ne=[7,1,5,3]","support","1","yes","Vulkan"
"Vulkan0","TRUNC","type=f32,ne=[1024,1024,1,1]","support","1","yes","Vulkan"
"Vulkan0","DIAG_MASK_INF","type=f32,ne=[10,10,1,1],n_past=5","support","1","yes","Vulkan"
"Vulkan0","DIAG_MASK_INF","type=f32,ne=[10,10,3,1],n_past=5","support","1","yes","Vulkan"
"Vulkan0","DIAG_MASK_INF","type=f32,ne=[10,10,3,2],n_past=5","support","1","yes","Vulkan"
@ -9411,28 +9448,405 @@
"Vulkan0","CONCAT","type=i32,ne_a=[11,12,13,14],ne_b_d=7,dim=2,v=3","support","1","yes","Vulkan"
"Vulkan0","CONCAT","type=f32,ne_a=[11,12,13,14],ne_b_d=7,dim=3,v=3","support","1","yes","Vulkan"
"Vulkan0","CONCAT","type=i32,ne_a=[11,12,13,14],ne_b_d=7,dim=3,v=3","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[3,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[4,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[7,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[8,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[15,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[16,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[31,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[32,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[63,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[64,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[127,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[128,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[255,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[256,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[511,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[512,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[1023,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[1024,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2047,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2048,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[4095,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[4096,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[8191,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[8192,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[16383,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[16384,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[32767,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[32768,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[65535,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[65536,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[131071,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[131072,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[262143,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[262144,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[524287,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[524288,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[1048575,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[1048576,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[16,10,10,10],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[60,10,10,10],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[1023,2,1,3],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[1024,2,1,3],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[1025,2,1,3],order=0","support","0","no","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[16384,1,1,1],order=0","support","0","no","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2047,2,1,3],order=0","support","0","no","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2048,2,1,3],order=0","support","0","no","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2049,2,1,3],order=0","support","0","no","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[1025,2,1,3],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2047,2,1,3],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2048,2,1,3],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2049,2,1,3],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2,8,8192,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[8,1,1,1],order=1","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[3,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[4,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[7,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[8,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[15,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[16,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[31,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[32,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[63,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[64,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[127,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[128,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[255,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[256,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[511,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[512,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[1023,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[1024,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2047,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2048,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[4095,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[4096,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[8191,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[8192,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[16383,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[16384,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[32767,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[32768,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[65535,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[65536,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[131071,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[131072,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[262143,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[262144,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[524287,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[524288,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[1048575,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[1048576,1,1,1],order=0","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[16,10,10,10],order=1","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[60,10,10,10],order=1","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[1023,2,1,3],order=1","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[1024,2,1,3],order=1","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[1025,2,1,3],order=1","support","0","no","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[16384,1,1,1],order=1","support","0","no","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2047,2,1,3],order=1","support","0","no","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2048,2,1,3],order=1","support","0","no","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2049,2,1,3],order=1","support","0","no","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[1025,2,1,3],order=1","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2047,2,1,3],order=1","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2048,2,1,3],order=1","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2049,2,1,3],order=1","support","1","yes","Vulkan"
"Vulkan0","ARGSORT","type=f32,ne=[2,8,8192,1],order=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[12,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[13,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[13,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[15,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[15,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[15,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[19,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[19,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[19,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[19,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[27,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[27,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[27,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[27,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[27,1,2,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[43,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[43,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[43,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[43,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[43,1,2,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[64,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[75,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[64,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[75,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[64,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[75,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[64,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[75,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[64,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[75,1,2,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[128,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[139,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[128,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[139,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[128,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[139,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[128,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[139,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[128,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[139,1,2,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[128,1,1,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[139,1,2,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[256,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[267,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[256,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[267,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[256,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[267,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[256,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[267,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[256,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[267,1,2,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[256,1,1,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[267,1,2,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[512,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[523,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[512,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[523,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[512,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[523,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[512,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[523,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[512,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[523,1,2,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[512,1,1,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[523,1,2,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[512,1,1,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[523,1,2,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1024,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1035,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1024,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1035,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1024,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1035,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1024,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1035,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1024,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1035,1,2,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1024,1,1,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1035,1,2,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1024,1,1,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1035,1,2,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1024,1,1,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1035,1,2,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2048,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2059,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2048,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2059,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2048,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2059,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2048,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2059,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2048,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2059,1,2,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2048,1,1,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2059,1,2,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2048,1,1,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2059,1,2,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2048,1,1,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2059,1,2,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4096,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4107,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4096,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4107,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4096,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4107,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4096,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4107,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4096,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4107,1,2,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4096,1,1,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4107,1,2,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4096,1,1,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4107,1,2,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4096,1,1,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[4107,1,2,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8192,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8203,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8192,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8203,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8192,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8203,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8192,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8203,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8192,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8203,1,2,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8192,1,1,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8203,1,2,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8192,1,1,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8203,1,2,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8192,1,1,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[8203,1,2,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16384,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16395,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16384,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16395,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16384,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16395,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16384,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16395,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16384,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16395,1,2,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16384,1,1,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16395,1,2,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16384,1,1,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16395,1,2,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16384,1,1,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16395,1,2,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16384,1,1,1],k=9999","support","0","no","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16395,1,2,1],k=9999","support","0","no","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32768,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32779,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32768,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32779,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32768,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32779,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32768,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32779,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32768,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32779,1,2,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32768,1,1,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32779,1,2,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32768,1,1,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32779,1,2,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32768,1,1,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32779,1,2,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32768,1,1,1],k=9999","support","0","no","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[32779,1,2,1],k=9999","support","0","no","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65536,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65547,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65536,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65547,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65536,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65547,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65536,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65547,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65536,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65547,1,2,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65536,1,1,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65547,1,2,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65536,1,1,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65547,1,2,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65536,1,1,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65547,1,2,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65536,1,1,1],k=9999","support","0","no","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[65547,1,2,1],k=9999","support","0","no","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131072,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131083,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131072,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131083,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131072,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131083,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131072,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131083,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131072,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131083,1,2,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131072,1,1,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131083,1,2,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131072,1,1,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131083,1,2,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131072,1,1,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131083,1,2,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131072,1,1,1],k=9999","support","0","no","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[131083,1,2,1],k=9999","support","0","no","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262144,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262155,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262144,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262155,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262144,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262155,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262144,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262155,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262144,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262155,1,2,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262144,1,1,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262155,1,2,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262144,1,1,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262155,1,2,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262144,1,1,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262155,1,2,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262144,1,1,1],k=9999","support","0","no","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[262155,1,2,1],k=9999","support","0","no","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524288,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524299,1,2,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524288,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524299,1,2,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524288,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524299,1,2,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524288,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524299,1,2,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524288,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524299,1,2,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524288,1,1,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524299,1,2,1],k=100","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524288,1,1,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524299,1,2,1],k=500","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524288,1,1,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524299,1,2,1],k=1023","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524288,1,1,1],k=9999","support","0","no","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[524299,1,2,1],k=9999","support","0","no","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16,10,10,10],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[60,10,10,10],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1023,2,1,3],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1024,2,1,3],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1025,2,1,3],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16384,1,1,1],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2047,2,1,3],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2048,2,1,3],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2049,2,1,3],k=1","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16,10,10,10],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[60,10,10,10],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1023,2,1,3],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1024,2,1,3],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1025,2,1,3],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16384,1,1,1],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2047,2,1,3],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2048,2,1,3],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2049,2,1,3],k=2","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16,10,10,10],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[60,10,10,10],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1023,2,1,3],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1024,2,1,3],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1025,2,1,3],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16384,1,1,1],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2047,2,1,3],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2048,2,1,3],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2049,2,1,3],k=3","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16,10,10,10],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[60,10,10,10],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1023,2,1,3],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1024,2,1,3],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1025,2,1,3],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16384,1,1,1],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2047,2,1,3],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2048,2,1,3],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2049,2,1,3],k=7","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16,10,10,10],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[60,10,10,10],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1023,2,1,3],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1024,2,1,3],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[1025,2,1,3],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[16384,1,1,1],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2047,2,1,3],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2048,2,1,3],k=15","support","1","yes","Vulkan"
"Vulkan0","TOP_K","type=f32,ne=[2049,2,1,3],k=15","support","1","yes","Vulkan"
"Vulkan0","UPSCALE","type=f32,ne=[512,512,3,2],scale_factor=2,mode=nearest,transpose=0","support","1","yes","Vulkan"
"Vulkan0","UPSCALE","type=f32,ne=[512,512,3,2],scale_factor=2,mode=nearest,transpose=1","support","1","yes","Vulkan"
"Vulkan0","UPSCALE","type=f32,ne=[2,5,7,11],ne_tgt=[5,7,11,13],mode=nearest,flags=none","support","1","yes","Vulkan"
@ -9445,6 +9859,10 @@
"Vulkan0","UPSCALE","type=f32,ne=[512,512,3,2],scale_factor=2,mode=bicubic,transpose=1","support","1","yes","Vulkan"
"Vulkan0","UPSCALE","type=f32,ne=[2,5,7,11],ne_tgt=[5,7,11,13],mode=bicubic,flags=none","support","1","yes","Vulkan"
"Vulkan0","UPSCALE","type=f32,ne=[5,7,11,13],ne_tgt=[2,5,7,11],mode=bicubic,flags=none","support","1","yes","Vulkan"
"Vulkan0","UPSCALE","type=f32,ne=[512,512,3,2],scale_factor=2,mode=513,transpose=0","support","0","no","Vulkan"
"Vulkan0","UPSCALE","type=f32,ne=[512,512,3,2],scale_factor=2,mode=513,transpose=1","support","0","no","Vulkan"
"Vulkan0","UPSCALE","type=f32,ne=[2,5,7,11],ne_tgt=[5,7,11,13],mode=bilinear,flags=none","support","0","no","Vulkan"
"Vulkan0","UPSCALE","type=f32,ne=[5,7,11,13],ne_tgt=[2,5,7,11],mode=bilinear,flags=none","support","0","no","Vulkan"
"Vulkan0","UPSCALE","type=f32,ne=[2,5,7,11],ne_tgt=[5,7,11,13],mode=bilinear,flags=align_corners","support","1","yes","Vulkan"
"Vulkan0","UPSCALE","type=f32,ne=[1,4,3,2],ne_tgt=[2,8,3,2],mode=bilinear,flags=align_corners","support","1","yes","Vulkan"
"Vulkan0","UPSCALE","type=f32,ne=[4,1,3,2],ne_tgt=[1,1,3,2],mode=bilinear,flags=align_corners","support","1","yes","Vulkan"
@ -9479,23 +9897,37 @@
"Vulkan0","PAD_REFLECT_1D","type=f32,ne_a=[3000,384,4,1],pad_0=10,pad_1=9","support","0","no","Vulkan"
"Vulkan0","ROLL","shift0=3,shift1=-2,shift3=1,shift4=-1","support","1","yes","Vulkan"
"Vulkan0","ARANGE","type=f32,start=0.000000,stop=10.000000,step=1.000000","support","1","yes","Vulkan"
"Vulkan0","ARANGE","type=f32,start=0.000000,stop=1048576.000000,step=1.000000","support","1","yes","Vulkan"
"Vulkan0","TIMESTEP_EMBEDDING","type=f32,ne_a=[2,1,1,1],dim=320,max_period=10000","support","1","yes","Vulkan"
"Vulkan0","LEAKY_RELU","type=f32,ne_a=[10,5,4,3],negative_slope=0.100000","support","1","yes","Vulkan"
"Vulkan0","CUMSUM","type=f32,ne=[10,5,4,3]","support","0","no","Vulkan"
"Vulkan0","CUMSUM","type=f32,ne=[10,5,4,3]","support","1","yes","Vulkan"
"Vulkan0","CUMSUM","type=f32,ne=[127,5,4,3]","support","1","yes","Vulkan"
"Vulkan0","CUMSUM","type=f32,ne=[128,5,4,3]","support","1","yes","Vulkan"
"Vulkan0","CUMSUM","type=f32,ne=[255,5,4,3]","support","1","yes","Vulkan"
"Vulkan0","CUMSUM","type=f32,ne=[256,5,4,3]","support","1","yes","Vulkan"
"Vulkan0","CUMSUM","type=f32,ne=[511,5,4,3]","support","1","yes","Vulkan"
"Vulkan0","CUMSUM","type=f32,ne=[512,5,4,3]","support","1","yes","Vulkan"
"Vulkan0","CUMSUM","type=f32,ne=[1023,5,4,3]","support","1","yes","Vulkan"
"Vulkan0","CUMSUM","type=f32,ne=[1024,5,4,3]","support","1","yes","Vulkan"
"Vulkan0","CUMSUM","type=f32,ne=[2047,5,4,3]","support","1","yes","Vulkan"
"Vulkan0","CUMSUM","type=f32,ne=[2048,5,4,3]","support","1","yes","Vulkan"
"Vulkan0","CUMSUM","type=f32,ne=[242004,1,1,1]","support","1","yes","Vulkan"
"Vulkan0","CUMSUM","type=f32,ne=[375960,1,1,1]","support","1","yes","Vulkan"
"Vulkan0","XIELU","type=f32,ne=[10,5,4,3]","support","0","no","Vulkan"
"Vulkan0","TRI","type=f32,ne=[10,10,4,3],tri_type=3","support","0","no","Vulkan"
"Vulkan0","TRI","type=f32,ne=[10,10,4,3],tri_type=2","support","0","no","Vulkan"
"Vulkan0","TRI","type=f32,ne=[10,10,4,3],tri_type=1","support","0","no","Vulkan"
"Vulkan0","TRI","type=f32,ne=[10,10,4,3],tri_type=0","support","0","no","Vulkan"
"Vulkan0","TRI","type=f32,ne=[10,10,4,3],tri_type=3","support","1","yes","Vulkan"
"Vulkan0","TRI","type=f32,ne=[10,10,4,3],tri_type=2","support","1","yes","Vulkan"
"Vulkan0","TRI","type=f32,ne=[10,10,4,3],tri_type=1","support","1","yes","Vulkan"
"Vulkan0","TRI","type=f32,ne=[10,10,4,3],tri_type=0","support","1","yes","Vulkan"
"Vulkan0","FILL","type=f32,ne=[10,10,4,3],c=0.000000","support","1","yes","Vulkan"
"Vulkan0","FILL","type=f32,ne=[303,207,11,3],c=2.000000","support","1","yes","Vulkan"
"Vulkan0","FILL","type=f32,ne=[800,600,4,4],c=-152.000000","support","1","yes","Vulkan"
"Vulkan0","SOLVE_TRI","type=f32,ne_lhs=[10,10,4,3],ne_rhs=[3,10,4,3]","support","0","no","Vulkan"
"Vulkan0","SOLVE_TRI","type=f32,ne_lhs=[11,11,1,1],ne_rhs=[5,11,1,1]","support","0","no","Vulkan"
"Vulkan0","SOLVE_TRI","type=f32,ne_lhs=[17,17,2,4],ne_rhs=[9,17,2,4]","support","0","no","Vulkan"
"Vulkan0","SOLVE_TRI","type=f32,ne_lhs=[30,30,7,1],ne_rhs=[8,30,7,1]","support","0","no","Vulkan"
"Vulkan0","SOLVE_TRI","type=f32,ne_lhs=[42,42,5,2],ne_rhs=[10,42,5,2]","support","0","no","Vulkan"
"Vulkan0","SOLVE_TRI","type=f32,ne_lhs=[64,64,2,2],ne_rhs=[10,64,2,2]","support","0","no","Vulkan"
"Vulkan0","FILL","type=f32,ne=[2048,512,2,2],c=3.500000","support","1","yes","Vulkan"
"Vulkan0","SOLVE_TRI","type=f32,ne_lhs=[10,10,4,3],ne_rhs=[3,10,4,3]","support","1","yes","Vulkan"
"Vulkan0","SOLVE_TRI","type=f32,ne_lhs=[11,11,1,1],ne_rhs=[5,11,1,1]","support","1","yes","Vulkan"
"Vulkan0","SOLVE_TRI","type=f32,ne_lhs=[17,17,2,4],ne_rhs=[9,17,2,4]","support","1","yes","Vulkan"
"Vulkan0","SOLVE_TRI","type=f32,ne_lhs=[30,30,7,1],ne_rhs=[8,30,7,1]","support","1","yes","Vulkan"
"Vulkan0","SOLVE_TRI","type=f32,ne_lhs=[42,42,5,2],ne_rhs=[10,42,5,2]","support","1","yes","Vulkan"
"Vulkan0","SOLVE_TRI","type=f32,ne_lhs=[64,64,2,2],ne_rhs=[10,64,2,2]","support","1","yes","Vulkan"
"Vulkan0","SOLVE_TRI","type=f32,ne_lhs=[100,100,4,4],ne_rhs=[41,100,4,4]","support","0","no","Vulkan"
"Vulkan0","PAD","type=f32,ne_a=[512,512,1,1],lp0=0,rp0=1,lp1=0,rp1=1,lp2=0,rp2=0,lp3=0,rp3=0,v=0","support","1","yes","Vulkan"
"Vulkan0","PAD","type=f32,ne_a=[11,22,33,44],lp0=1,rp0=2,lp1=3,rp1=4,lp2=5,rp2=6,lp3=7,rp3=8,v=0","support","1","yes","Vulkan"

Can't render this file because it is too large.

18741
docs/ops/WebGPU.csv Normal file
View File

File diff suppressed because it is too large Load Diff

18741
docs/ops/ZenDNN.csv Normal file
View File

File diff suppressed because it is too large Load Diff

1
examples/CMakeLists.txt
View File

@ -20,6 +20,7 @@ else()
add_subdirectory(gguf-hash)
add_subdirectory(gguf)
add_subdirectory(idle)
add_subdirectory(lookahead)
add_subdirectory(lookup)
add_subdirectory(parallel)

30
examples/batched/batched.cpp
View File

@ -2,6 +2,7 @@
#include "common.h"
#include "log.h"
#include "llama.h"
#include "sampling.h"
#include <algorithm>
#include <cstdio>
@ -64,17 +65,23 @@ int main(int argc, char ** argv) {
ctx_params.n_ctx = n_kv_req;
ctx_params.n_batch = std::max(n_predict, n_parallel);
llama_context * ctx = llama_init_from_model(model, ctx_params);
auto sparams = llama_sampler_chain_default_params();
sparams.no_perf = false;
llama_sampler * smpl = llama_sampler_chain_init(sparams);
std::vector<llama_sampler *> samplers;
llama_sampler_chain_add(smpl, llama_sampler_init_top_k(params.sampling.top_k));
llama_sampler_chain_add(smpl, llama_sampler_init_top_p(params.sampling.top_p, params.sampling.min_keep));
llama_sampler_chain_add(smpl, llama_sampler_init_temp (params.sampling.temp));
llama_sampler_chain_add(smpl, llama_sampler_init_dist (params.sampling.seed));
for (int32_t i = 0; i < n_parallel; ++i) {
llama_sampler * smpl = llama_sampler_chain_init(sparams);
llama_sampler_chain_add(smpl, llama_sampler_init_top_k(params.sampling.top_k));
llama_sampler_chain_add(smpl, llama_sampler_init_top_p(params.sampling.top_p, params.sampling.min_keep));
llama_sampler_chain_add(smpl, llama_sampler_init_temp (params.sampling.temp));
llama_sampler_chain_add(smpl, llama_sampler_init_dist (params.sampling.seed));
samplers.push_back(smpl);
}
llama_context * ctx = llama_init_from_model(model, ctx_params);
if (ctx == NULL) {
LOG_ERR("%s: error: failed to create the llama_context\n" , __func__);
@ -173,7 +180,7 @@ int main(int argc, char ** argv) {
continue;
}
const llama_token new_token_id = llama_sampler_sample(smpl, ctx, i_batch[i]);
const llama_token new_token_id = llama_sampler_sample(samplers[i], ctx, i_batch[i]);
// is it an end of generation? -> mark the stream as finished
if (llama_vocab_is_eog(vocab, new_token_id) || n_cur == n_predict) {
@ -229,14 +236,17 @@ int main(int argc, char ** argv) {
__func__, n_decode, (t_main_end - t_main_start) / 1000000.0f, n_decode / ((t_main_end - t_main_start) / 1000000.0f));
LOG("\n");
llama_perf_sampler_print(smpl);
llama_perf_sampler_print(samplers[0]);
llama_perf_context_print(ctx);
fprintf(stderr, "\n");
llama_batch_free(batch);
llama_sampler_free(smpl);
for (auto & sampler_config : samplers) {
llama_sampler_free(sampler_config);
}
llama_free(ctx);
llama_model_free(model);

6
examples/embedding/embedding.cpp
View File

@ -131,10 +131,10 @@ int main(int argc, char ** argv) {
llama_numa_init(params.numa);
// load the model
common_init_result llama_init = common_init_from_params(params);
auto llama_init = common_init_from_params(params);
llama_model * model = llama_init.model.get();
llama_context * ctx = llama_init.context.get();
auto * model = llama_init->model();
auto * ctx = llama_init->context();
if (model == NULL) {
LOG_ERR("%s: unable to load model\n", __func__);

6
examples/eval-callback/eval-callback.cpp
View File

@ -202,10 +202,10 @@ int main(int argc, char ** argv) {
params.warmup = false;
// init
common_init_result llama_init = common_init_from_params(params);
auto llama_init = common_init_from_params(params);
llama_model * model = llama_init.model.get();
llama_context * ctx = llama_init.context.get();
auto * model = llama_init->model();
auto * ctx = llama_init->context();
if (model == nullptr || ctx == nullptr) {
LOG_ERR("%s : failed to init\n", __func__);

11
examples/gen-docs/gen-docs.cpp
View File

@ -14,12 +14,13 @@ static void write_table_header(std::ofstream & file) {
static void write_table_entry(std::ofstream & file, const common_arg & opt) {
file << "| `";
// args
for (const auto & arg : opt.args) {
if (arg == opt.args.front()) {
auto all_args = opt.get_args();
for (const auto & arg : all_args) {
if (arg == all_args.front()) {
file << arg;
if (opt.args.size() > 1) file << ", ";
if (all_args.size() > 1) file << ", ";
} else {
file << arg << (arg != opt.args.back() ? ", " : "");
file << arg << (arg != all_args.back() ? ", " : "");
}
}
// value hint
@ -76,7 +77,7 @@ static void export_md(std::string fname, llama_example ex) {
}
int main(int, char **) {
export_md("autogen-main.md", LLAMA_EXAMPLE_MAIN);
export_md("autogen-main.md", LLAMA_EXAMPLE_COMPLETION);
export_md("autogen-server.md", LLAMA_EXAMPLE_SERVER);
return 0;

5
examples/idle/CMakeLists.txt Normal file
View File

@ -0,0 +1,5 @@
set(TARGET llama-idle)
add_executable(${TARGET} idle.cpp)
install(TARGETS ${TARGET} RUNTIME)
target_link_libraries(${TARGET} PRIVATE llama common ${CMAKE_THREAD_LIBS_INIT})
target_compile_features(${TARGET} PRIVATE cxx_std_11)

3
examples/idle/README.md Normal file
View File

@ -0,0 +1,3 @@
# llama.cpp/example/idle
https://github.com/ggml-org/llama.cpp/pull/17766

110
examples/idle/idle.cpp Normal file
View File

@ -0,0 +1,110 @@
#include "arg.h"
#include "common.h"
#include "log.h"
#include "llama.h"
#include <cmath>
#include <cstdio>
#include <cstring>
#include <string>
#include <thread>
#include <vector>
static void print_usage(int /*argc*/, char ** argv) {
printf("\nexample usage:\n");
printf("\n %s -m model.gguf [-ngl n_gpu_layers]\n", argv[0]);
printf("\n");
}
int main(int argc, char ** argv) {
common_params params;
if (!common_params_parse(argc, argv, params, LLAMA_EXAMPLE_COMMON, print_usage)) {
return 1;
}
common_init();
// init LLM
llama_backend_init();
llama_numa_init(params.numa);
// initialize the model
llama_model_params model_params = common_model_params_to_llama(params);
llama_model * model = llama_model_load_from_file(params.model.path.c_str(), model_params);
if (model == NULL) {
LOG_ERR("%s: error: unable to load model\n" , __func__);
return 1;
}
const llama_vocab * vocab = llama_model_get_vocab(model);
// we need just a dummy token to evaluate
std::vector<llama_token> prompt_tokens(1, llama_vocab_bos(vocab));
llama_context_params ctx_params = llama_context_default_params();
ctx_params.n_ctx = 512;
ctx_params.n_batch = 512;
ctx_params.no_perf = false;
llama_context * ctx = llama_init_from_model(model, ctx_params);
if (ctx == NULL) {
fprintf(stderr , "%s: error: failed to create the llama_context\n" , __func__);
return 1;
}
llama_batch batch = llama_batch_get_one(prompt_tokens.data(), prompt_tokens.size());
const int n_iters = 3;
// warm-up
llama_decode(ctx, batch);
llama_memory_clear(llama_get_memory(ctx), true);
llama_synchronize(ctx);
for (int64_t t_pause_ms = 0; t_pause_ms <= 4000; t_pause_ms += 800) {
double t_sum_us = 0.0;
double t_sum2_us = 0.0;
for (int i = 0; i < n_iters; i++) {
// this pause is important - it simulates "idle GPU"
std::this_thread::sleep_for(std::chrono::milliseconds(t_pause_ms));
const int64_t t_start_us = llama_time_us();
// this should take constant time
llama_decode(ctx, batch);
llama_synchronize(ctx);
const int64_t t_end_us = llama_time_us();
const double t_cur_us = t_end_us - t_start_us;
#if 1
// print individual decode times
printf(" - decode time: %8.2f ms\n", t_cur_us / 1000);
#endif
t_sum_us += t_cur_us;
t_sum2_us += t_cur_us * t_cur_us;
llama_memory_clear(llama_get_memory(ctx), true);
llama_synchronize(ctx); // just in case
}
const double t_avg_us = t_sum_us / n_iters;
const double t_dev_us = sqrt((t_sum2_us / (n_iters - 1)) - (t_avg_us * t_avg_us * n_iters) / (n_iters - 1));
printf("iters: %4d, pause: %5d ms, avg decode time: %8.2f +/- %4.2f ms\n", n_iters, (int) t_pause_ms, t_avg_us / 1000, t_dev_us / 1000);
fflush(stdout);
}
llama_free(ctx);
llama_model_free(model);
return 0;
}

4
examples/json_schema_to_grammar.py
View File

@ -231,9 +231,9 @@ DOT = '[^\\x0A\\x0D]'
RESERVED_NAMES = set(["root", "dot", *PRIMITIVE_RULES.keys(), *STRING_FORMAT_RULES.keys()])
INVALID_RULE_CHARS_RE = re.compile(r'[^a-zA-Z0-9-]+')
GRAMMAR_LITERAL_ESCAPE_RE = re.compile(r'[\r\n"]')
GRAMMAR_LITERAL_ESCAPE_RE = re.compile(r'[\r\n"\\]')
GRAMMAR_RANGE_LITERAL_ESCAPE_RE = re.compile(r'[\r\n"\]\-\\]')
GRAMMAR_LITERAL_ESCAPES = {'\r': '\\r', '\n': '\\n', '"': '\\"', '-': '\\-', ']': '\\]'}
GRAMMAR_LITERAL_ESCAPES = {'\r': '\\r', '\n': '\\n', '"': '\\"', '-': '\\-', ']': '\\]', '\\': '\\\\'}
NON_LITERAL_SET = set('|.()[]{}*+?')
ESCAPED_IN_REGEXPS_BUT_NOT_IN_LITERALS = set('^$.[]()|{}*+?')

6
examples/lookahead/lookahead.cpp
View File

@ -55,10 +55,10 @@ int main(int argc, char ** argv) {
llama_numa_init(params.numa);
// load the target model
common_init_result llama_init = common_init_from_params(params);
auto llama_init = common_init_from_params(params);
llama_model * model = llama_init.model.get();
llama_context * ctx = llama_init.context.get();
auto * model = llama_init->model();
auto * ctx = llama_init->context();
auto * mem = llama_get_memory(ctx);

8
examples/lookup/lookup-create.cpp
View File

@ -18,16 +18,16 @@ int main(int argc, char ** argv){
llama_numa_init(params.numa);
// load the model
common_init_result llama_init = common_init_from_params(params);
auto llama_init = common_init_from_params(params);
llama_model_ptr & model = llama_init.model;
llama_context_ptr & ctx = llama_init.context;
auto * model = llama_init->model();
auto * ctx = llama_init->context();
GGML_ASSERT(model != nullptr);
// tokenize the prompt
std::vector<llama_token> inp;
inp = common_tokenize(ctx.get(), params.prompt, true, true);
inp = common_tokenize(ctx, params.prompt, true, true);
fprintf(stderr, "%s: tokenization done\n", __func__);
common_ngram_cache ngram_cache;

8
examples/lookup/lookup-stats.cpp
View File

@ -28,13 +28,13 @@ int main(int argc, char ** argv){
llama_numa_init(params.numa);
// load the model
common_init_result llama_init = common_init_from_params(params);
auto llama_init = common_init_from_params(params);
llama_context_ptr & ctx = llama_init.context;
llama_context * ctx = llama_init->context();
// tokenize the prompt
std::vector<llama_token> inp;
inp = common_tokenize(ctx.get(), params.prompt, true, true);
inp = common_tokenize(ctx, params.prompt, true, true);
common_ngram_cache ngram_cache_context;
common_ngram_cache ngram_cache_dynamic;
@ -65,7 +65,7 @@ int main(int argc, char ** argv){
}
const int n_input = inp.size();
const int n_ctx = llama_n_ctx(ctx.get());
const int n_ctx = llama_n_ctx(ctx);
int n_drafted = 0;
int n_accept = 0;

6
examples/lookup/lookup.cpp
View File

@ -29,10 +29,10 @@ int main(int argc, char ** argv){
llama_numa_init(params.numa);
// load the model
common_init_result llama_init = common_init_from_params(params);
auto llama_init = common_init_from_params(params);
llama_model * model = llama_init.model.get();
llama_context * ctx = llama_init.context.get();
auto * model = llama_init->model();
auto * ctx = llama_init->context();
const llama_vocab * vocab = llama_model_get_vocab(model);

2
examples/model-conversion/logits.cpp
View File

@ -144,7 +144,7 @@ int main(int argc, char ** argv) {
return 1;
}
std::string s(buf, n);
printf("%s", s.c_str());
printf("%s (%d)", s.c_str(), id);
}
printf("\n");

28
examples/model-conversion/scripts/causal/compare-logits.py
View File

@ -1,10 +1,13 @@
#!/usr/bin/env python3
import numpy as np
import sys
import os
import numpy as np
from pathlib import Path
# Add utils directory to path for direct script execution
sys.path.insert(0, str(Path(__file__).parent.parent / "utils"))
from common import get_model_name_from_env_path # type: ignore[import-not-found]
def quick_logits_check(pytorch_file, llamacpp_file):
"""Lightweight sanity check before NMSE"""
@ -32,27 +35,16 @@ def quick_logits_check(pytorch_file, llamacpp_file):
print(f"Top 10 llama.cpp logits: {llamacpp_logits[llamacpp_top10]}")
print(f"Max absolute difference: {max_diff:.4f}")
if max_diff > 1.0:
print(f"❌ NOK: Large differences detected - max diff: {max_diff:.4f}")
return False
return True
def main():
model_path = os.getenv('MODEL_PATH')
if not model_path:
print("Error: MODEL_PATH environment variable not set")
sys.exit(1)
if not os.path.exists(model_path):
print(f"Error: Model file not found: {model_path}")
sys.exit(1)
model_name = os.path.basename(model_path)
model_name = get_model_name_from_env_path('MODEL_PATH')
data_dir = Path("data")
pytorch_file = data_dir / f"pytorch-{model_name}.bin"
llamacpp_file = data_dir / f"llamacpp-{model_name}.bin"
llamacpp_model_name = get_model_name_from_env_path('CONVERTED_MODEL')
print(f"Using converted model: {llamacpp_model_name}")
llamacpp_file = data_dir / f"llamacpp-{llamacpp_model_name}.bin"
if not pytorch_file.exists():
print(f"Error: PyTorch logits file not found: {pytorch_file}")

2
examples/model-conversion/scripts/causal/run-org-model.py
View File

@ -200,7 +200,7 @@ with torch.no_grad():
logits = outputs.logits
# Extract logits for the last token (next token prediction)
last_logits = logits[0, -1, :].cpu().numpy()
last_logits = logits[0, -1, :].float().cpu().numpy()
print(f"Logits shape: {logits.shape}")
print(f"Last token logits shape: {last_logits.shape}")

0
examples/model-conversion/scripts/utils/__init__.py Normal file
View File

7
examples/model-conversion/scripts/utils/check-nmse.py
View File

@ -5,6 +5,7 @@ import sys
import os
import argparse
from pathlib import Path
from common import get_model_name_from_env_path # type: ignore[import-not-found]
def calculate_nmse(reference, test):
mse = np.mean((test - reference) ** 2)
@ -67,11 +68,13 @@ def main():
parser.add_argument('-m', '--model-path', required=True, help='Path to the model directory')
args = parser.parse_args()
model_name = os.path.basename(args.model_path)
model_name = get_model_name_from_env_path('MODEL_PATH')
data_dir = Path("data")
pytorch_file = data_dir / f"pytorch-{model_name}.bin"
llamacpp_file = data_dir / f"llamacpp-{model_name}.bin"
llamacpp_model_name = get_model_name_from_env_path('CONVERTED_MODEL')
llamacpp_file = data_dir / f"llamacpp-{llamacpp_model_name}.bin"
print(f"Model name: {model_name}")
print(f"PyTorch logits file: {pytorch_file}")

20
examples/model-conversion/scripts/utils/common.py Normal file
View File

@ -0,0 +1,20 @@
#!/usr/bin/env python3
import os
import sys
def get_model_name_from_env_path(env_path_name):
model_path = os.getenv(env_path_name)
if not model_path:
print(f"Error: {env_path_name} environment variable not set")
sys.exit(1)
if not os.path.exists(model_path):
print(f"Error: Model file not found: {model_path}")
sys.exit(1)
name = os.path.basename(os.path.normpath(model_path))
if name.endswith(".gguf"):
name = name[:-5]
return name

6
examples/parallel/parallel.cpp
View File

@ -192,10 +192,10 @@ int main(int argc, char ** argv) {
llama_numa_init(params.numa);
// load the target model
common_init_result llama_init = common_init_from_params(params);
auto llama_init = common_init_from_params(params);
llama_model * model = llama_init.model.get();
llama_context * ctx = llama_init.context.get();
auto * model = llama_init->model();
auto * ctx = llama_init->context();
auto * mem = llama_get_memory(ctx);

6
examples/retrieval/retrieval.cpp
View File

@ -149,10 +149,10 @@ int main(int argc, char ** argv) {
llama_numa_init(params.numa);
// load the model
common_init_result llama_init = common_init_from_params(params);
auto llama_init = common_init_from_params(params);
llama_model * model = llama_init.model.get();
llama_context * ctx = llama_init.context.get();
auto * model = llama_init->model();
auto * ctx = llama_init->context();
if (model == NULL) {
LOG_ERR("%s: unable to load model\n", __func__);

12
examples/save-load-state/save-load-state.cpp
View File

@ -34,10 +34,10 @@ int main(int argc, char ** argv) {
std::string result2;
// init
common_init_result llama_init = common_init_from_params(params);
auto llama_init = common_init_from_params(params);
llama_model * model = llama_init.model.get();
llama_context * ctx = llama_init.context.get();
auto * model = llama_init->model();
auto * ctx = llama_init->context();
if (model == nullptr || ctx == nullptr) {
fprintf(stderr, "%s : failed to init\n", __func__);
@ -241,6 +241,12 @@ int main(int argc, char ** argv) {
llama_batch_free(batch);
// this one is managed by common_init_result
//llama_free(ctx);
llama_free(ctx2);
llama_free(ctx3);
if (result0 != result2) {
fprintf(stderr, "\n%s : error : the seq restore generation is different\n", __func__);
return 1;

1
examples/simple-cmake-pkg/README.md
View File

@ -18,6 +18,7 @@ cd llama.cpp
cmake -S . -B build
cmake --build build
cmake --install build --prefix inst
```
### Build simple-cmake-pkg

14
examples/speculative-simple/speculative-simple.cpp
View File

@ -40,10 +40,10 @@ int main(int argc, char ** argv) {
llama_context * ctx_dft = NULL;
// load the target model
common_init_result llama_init_tgt = common_init_from_params(params);
auto llama_init_tgt = common_init_from_params(params);
model_tgt = llama_init_tgt.model.get();
ctx_tgt = llama_init_tgt.context.get();
model_tgt = llama_init_tgt->model();
ctx_tgt = llama_init_tgt->context();
const llama_vocab * vocab = llama_model_get_vocab(model_tgt);
@ -61,10 +61,10 @@ int main(int argc, char ** argv) {
params.cpuparams_batch.n_threads = params.speculative.cpuparams_batch.n_threads;
params.tensor_buft_overrides = params.speculative.tensor_buft_overrides;
common_init_result llama_init_dft = common_init_from_params(params);
auto llama_init_dft = common_init_from_params(params);
//model_dft = llama_init_dft.model.get();
ctx_dft = llama_init_dft.context.get();
//model_dft = llama_init_dft->model();
ctx_dft = llama_init_dft->context();
if (!common_speculative_are_compatible(ctx_tgt, ctx_dft)) {
LOG_INF("the draft model '%s' is not compatible with the target model '%s'. tokens will be translated between the draft and target models.\n", params.speculative.model.path.c_str(), params.model.path.c_str());
@ -255,6 +255,8 @@ int main(int argc, char ** argv) {
LOG_INF("target:\n\n");
common_perf_print(ctx_tgt, smpl);
llama_batch_free(batch_tgt);
common_sampler_free(smpl);
common_speculative_free(spec);

16
examples/speculative/speculative.cpp
View File

@ -71,10 +71,10 @@ int main(int argc, char ** argv) {
llama_context * ctx_dft = NULL;
// load the target model
common_init_result llama_init_tgt = common_init_from_params(params);
auto llama_init_tgt = common_init_from_params(params);
model_tgt = llama_init_tgt.model.get();
ctx_tgt = llama_init_tgt.context.get();
model_tgt = llama_init_tgt->model();
ctx_tgt = llama_init_tgt->context();
// load the draft model
params.devices = params.speculative.devices;
@ -87,10 +87,10 @@ int main(int argc, char ** argv) {
params.cpuparams_batch.n_threads = params.speculative.cpuparams_batch.n_threads;
params.tensor_buft_overrides = params.speculative.tensor_buft_overrides;
common_init_result llama_init_dft = common_init_from_params(params);
auto llama_init_dft = common_init_from_params(params);
model_dft = llama_init_dft.model.get();
ctx_dft = llama_init_dft.context.get();
model_dft = llama_init_dft->model();
ctx_dft = llama_init_dft->context();
const llama_vocab * vocab_tgt = llama_model_get_vocab(model_tgt);
const llama_vocab * vocab_dft = llama_model_get_vocab(model_dft);
@ -242,7 +242,7 @@ int main(int argc, char ** argv) {
bool accept = false;
if (params.sampling.temp > 0) {
// stochastic verification
common_sampler_sample(smpl, ctx_tgt, drafts[s_keep].i_batch_tgt[i_dft], true);
common_sampler_sample(smpl, ctx_tgt, drafts[s_keep].i_batch_tgt[i_dft]);
auto & dist_tgt = *common_sampler_get_candidates(smpl, true);
@ -491,7 +491,7 @@ int main(int argc, char ** argv) {
continue;
}
common_sampler_sample(drafts[s].smpl, ctx_dft, drafts[s].i_batch_dft, true);
common_sampler_sample(drafts[s].smpl, ctx_dft, drafts[s].i_batch_dft);
const auto * cur_p = common_sampler_get_candidates(drafts[s].smpl, true);

3
examples/sycl/run-llama2.sh
View File

@ -15,6 +15,9 @@ MODEL_FILE=models/llama-2-7b.Q4_0.gguf
NGL=99
CONTEXT=4096
#support malloc device memory more than 4GB.
export UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS=1
if [ $# -gt 0 ]; then
GGML_SYCL_DEVICE=$1
echo "use $GGML_SYCL_DEVICE as main GPU"

9
examples/sycl/run-llama3.sh
View File

@ -6,7 +6,7 @@
# If you want more control, DPC++ Allows selecting a specific device through the
# following environment variable
#export ONEAPI_DEVICE_SELECTOR="level_zero:0"
export ONEAPI_DEVICE_SELECTOR="level_zero:0"
source /opt/intel/oneapi/setvars.sh
#export GGML_SYCL_DEBUG=1
@ -18,11 +18,14 @@ MODEL_FILE=models/Meta-Llama-3.1-8B-Instruct-Q4_K_M.gguf
NGL=99 # Layers offloaded to the GPU. If the device runs out of memory, reduce this value according to the model you are using.
CONTEXT=4096
#support malloc device memory more than 4GB.
export UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS=1
if [ $# -gt 0 ]; then
GGML_SYCL_DEVICE=$1
echo "Using $GGML_SYCL_DEVICE as the main GPU"
ZES_ENABLE_SYSMAN=1 ./build/bin/llama-cli -m ${MODEL_FILE} -p "${INPUT_PROMPT}" -n 400 -e -ngl ${NGL} -c ${CONTEXT} -mg $GGML_SYCL_DEVICE -sm none
ZES_ENABLE_SYSMAN=1 ./build/bin/llama-cli -m ${MODEL_FILE} -p "${INPUT_PROMPT}" -n 400 -e -ngl ${NGL} -s 0 -c ${CONTEXT} -mg $GGML_SYCL_DEVICE -sm none
else
#use multiple GPUs with same max compute units
ZES_ENABLE_SYSMAN=1 ./build/bin/llama-cli -m ${MODEL_FILE} -p "${INPUT_PROMPT}" -n 400 -e -ngl ${NGL} -c ${CONTEXT}
ZES_ENABLE_SYSMAN=1 ./build/bin/llama-cli -m ${MODEL_FILE} -p "${INPUT_PROMPT}" -n 400 -e -ngl ${NGL} -s 0 -c ${CONTEXT}
fi

2
examples/sycl/win-run-llama2.bat
View File

@ -5,5 +5,7 @@
set INPUT2="Building a website can be done in 10 simple steps:\nStep 1:"
@call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" intel64 --force
:: support malloc device memory more than 4GB.
set UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS=1
.\build\bin\llama-cli.exe -m models\llama-2-7b.Q4_0.gguf -p %INPUT2% -n 400 -e -ngl 99 -s 0

4
examples/sycl/win-run-llama3.bat
View File

@ -5,5 +5,7 @@
set INPUT2="Building a website can be done in 10 simple steps:\nStep 1:"
@call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" intel64 --force
:: support malloc device memory more than 4GB.
set UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS=1
.\build\bin\llama-cli.exe -m models\Meta-Llama-3.1-8B-Instruct-Q4_K_M.gguf -p %INPUT2% -n 400 -e -ngl 99
.\build\bin\llama-cli.exe -m models\Meta-Llama-3.1-8B-Instruct-Q4_K_M.gguf -p %INPUT2% -n 400 -s 0 -e -ngl 99

Some files were not shown because too many files have changed in this diff Show More