When using group query attention, we have one workgroup per KV batch and this
can be very few workgroups (e.g. just 8 in some models). Enable split_k to
spread the work across SMs. This helps a lot when the KV cache is large.
* (wip) refactor downloading system [no ci]
* fix all examples
* fix mmproj with -hf
* gemma3: update readme
* only handle mmproj in llava example
* fix multi-shard download
* windows: fix problem with std::min and std::max
* fix 2
* ggml : FA with different K, V head sizes (CPU)
ggml-ci
* metal : add FA with HS=192
* metal : extend FA to support different K and V head sizes
ggml-ci
* metal : add FA vector kernels for heads K 192 and V 128
ggml-ci
* ggml : restrict op on other backends to equal head sizes
ggml-ci
* metal : optimize FA-vec kernel
ggml-ci
* metal : FA remove mq registers
* metal : improve MoE mul_mat_id condition
ggml-ci
* metal : fix comments + remove unnecessary addition
ggml-ci
* metal : avoid too much shared memory usage with mul_mat_id
ggml-ci
The OOB calculation could be wrong if the last iteration was during one of
the unrolled loops. Adjust the unrolling counts to avoid this. Add a couple
new backend tests that hit this failure on NVIDIA GPUs.
* tests: add mul_mat perf/functional tests for p021/nc vulkan shaders
* vulkan: Optimize mul_mat_vec p021 and nc shaders.
These shaders are used in attention calculations, and when the KV cache grows
large they start to dominate the run time. For the nc shader (which is called
with large 'k' dimension), use unrolling and vector loads. For the p021 shader
(which is called with large 'm' and small 'k' dimensions), take advantage of
grouped query attention to reuse loads from the A matrix for the whole group,
and reduce the number of workgroups (too much overhead from tiny dispatches).
Using subgroupAdd in the p021 shader also helps, use that conditionally.
- Find out active blocks per SM using cudaOccupancyMaxActiveBlocksPerMultiprocessor API. Use this value to determine the optimal parallel_blocks value.
- Prefer vector flash attention kernels over MMA kernel for BS=1
Fixes Issue: #12182
---------
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* sampler: turn lazy grammar trigger words to regexes
* add scripts/tool_bench.sh & .py
* constrain llama json output regardless of function name if matches at beginning
* update relaxed newline space rule in grammar tests
* support add_generation_prompt query parameter (useful for /apply_template)
* Update src/llama-grammar.cpp
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Support fp16 unary operations in the CUDA backend
* cpu: increase fp16 support for unary operators in the CPU backend
* cuda: increase fp16 support for unary operators in the CUDA backend
* Add test cases for fp16 unary operators
* metal: update supports_op for unary operators that don't support fp16, to prevent test-backend-ops from failing
* metal: fix PR comments for unary op support after fp16 unary tests
* Support float16-to-float16 add/sub/mul/div operations in the CUDA backend
* Add fp16 support for add/sub/mul/div on the CPU backend
* Add test cases for fp16 add/sub/mul/div
* Upgrade init_tensor API to return a ggml_status
To prepare for an 'abort-free' ggml
(ggml not to abort on OOMs but return a OOM status),
as agreeed with Diego in the ggml repo,
upgrade the init_tensor() and view_init() APIs
to return a ggml_status.
* misc fixes
---------
Co-authored-by: slaren <slarengh@gmail.com>
* extract & return thoughts in reasoning_content field (unless --reasoning-format) for DeepSeek R1 & Command R7B
* tool-calls: add deepseek r1 template (models/templates/llama-cpp-deepseek-r1.jinja) + hackommodate broken official template
* tool-calls: accommodate variety of wrong tool call opening tags both R1 Qwen 32B and 7B distills like to spit out
* server/oai: ensure content is null when there are tool calls, and reasoning_content appears before content for readability
* tool-calls: add DeepSeek R1 Qwen distills to server/README.md & server tests
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* initial porting of previous LLG patch
* update for new APIs
* build: integrate llguidance as an external project
* use '%llguidance' as marker to enable llg lark syntax
* add some docs
* clarify docs
* code style fixes
* remove llguidance.h from .gitignore
* fix tests when llg is enabled
* pass vocab not model to llama_sampler_init_llg()
* copy test-grammar-integration.cpp to test-llguidance.cpp
* clang fmt
* fix ref-count bug
* build and run test
* gbnf -> lark syntax
* conditionally include llguidance test based on LLAMA_LLGUIDANCE flag
* rename llguidance test file to test-grammar-llguidance.cpp
* add gh action for llg test
* align tests with LLG grammar syntax and JSON Schema spec
* llama_tokenizer() in fact requires valid utf8
* update llg
* format file
* add $LLGUIDANCE_LOG_LEVEL support
* fix whitespace
* fix warning
* include <cmath> for INFINITY
* add final newline
* fail llama_sampler_init_llg() at runtime
* Link gbnf_to_lark.py script; fix links; refer to llg docs for lexemes
* simplify #includes
* improve doc string for LLAMA_LLGUIDANCE
* typo in merge
* bump llguidance to 0.6.12
* add glm edge chat model
* use config partial_rotary_factor as rope ratio
* support for glm edge model
* vision model support
* remove debug info
* fix format
* llava.cpp trailing whitespace
* remove unused AutoTokenizer
* Update src/llama.cpp for not contain <|end|> or </s>
Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
* add edge template
* fix chat template
* fix confict
* fix confict
* fix ci err
* fix format err
* fix template err
* 9b hf chat support
* format
* format clip.cpp
* fix format
* Apply suggestions from code review
* Apply suggestions from code review
* Update examples/llava/clip.cpp
* fix format
* minor : style
---------
Co-authored-by: liyuhang <yuhang.li@zhipuai.cn>
Co-authored-by: piDack <pcdack@hotmail.co>
Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
Co-authored-by: liyuhang <yuhang.li@aminer.cn>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
---------
Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
Implemented ggml_sycl_op_soft_max() F16 src1(mask) support for which a pragma deprecation warning was added during #5021.
To do this, had to decouple it from ggml_sycl_op_flatten which always considered src1 to be of fp32 type(many OP functions are dependent on it).
* SYCL: SOFTMAX F16 mask support and other fixes
* test-backend-ops: Add F16 mask test cases
Now that we have batched mat-vec mul Vulkan shaders for up to n==8,
these tests weren't actually exercising the mat-mat mul path. Test
n==9 as well. Also, change to use all_types.
Add code similar to mul_mm_cm2 to force alignment of strides, to avoid
a performance regression.
Add noncontiguous FA tests in test-backend-ops.
Fixes#11268.
* vulkan: support copy from f32 to q4_0/q4_1/q5_0/q5_1/q8_0/iq4_nl
Shaders are based on cpy.cu.
* vulkan: support copy from q4_0/q4_1/q5_0/q5_1/q8_0/iq4_nl to f32
* ggml: copy q->f32 assumes some contiguity in the destination
* GGUF: C++ refactor, backend support, misc fixes
remove ggml_tensor.backend
update CODEOWNERS [no ci]
remove gguf_get_data from API
revise GGUF API data types
Make the mul_mat_vec shaders support N>1 (as a spec constant, NUM_COLS) where
the batch_strides are overloaded to hold the row strides. Put the loads from the
B matrix in the innermost loop because it should cache better.
Share some code for reducing the result values to memory in mul_mat_vec_base.
* tests: Add im2col perf tests
* vulkan: optimize im2col, more elements per thread
* vulkan: increase small tile size for NV_coopmat2
* vulkan: change im2col to 512 elements per workgroup
* sampling : refactor + optimize penalties sampler
ggml-ci
* common : apply ignore_eos as logit bias
ggml-ci
* batched : remove penalties sampler
* params : allow penalty_last_n == -1 to be equal to context size
ggml-ci
* common : by default, move the penalties at the end of the sampling chain
ggml-ci
* common : ignore all EOG tokens
Co-authored-by: Diego Devesa <slarengh@gmail.com>
* common : move back the penalties at the front of the sampling chain
ggml-ci
* readme : restore hint about --ignore-eos flag [no ci]
* llama : minor
ggml-ci
* webui : update
---------
Co-authored-by: Diego Devesa <slarengh@gmail.com>
* Add deepseek v1 arch & gigachat template
* improve template code
* add readme
* delete comments
* remove comment
* fix format
* lint llama.cpp
* fix order of deepseek and deepseek2, move gigachat temlate to the end of func
* fix order of deepseek and deepseek2 in constants; mark shared exp as deepseek arch need
* remove comments
* move deepseek above deepseek2
* change placement of gigachat chat template
* ggml_pad_reflect_1d defined in header
* implemented on CPU
* called the forward pass
* impl Metal kernel
* added Metal kernel
* added OP_PAD_REFLECT_1D in test-backend-ops.cpp
* add test-pad-reflect-1d test case
* test case support multiple backend
* llama : add enum for supported chat templates
* use "built-in" instead of "supported"
* arg: print list of built-in templates
* fix test
* update server README
* Templates: `mistral-v1`, `mistral-v2`, `mistral-v3`, `mistral-v3-tekken`
* Changed system message logic and added tests for all 4
* Invalid `system_message` instead of `content` fixed
* Removed tab-indented lines
* Added template code and test for `mistral-v7`
* Added all tests. Fixed bug with `tmpl == "llama2"` test.
* Replaced tabs with spaces.
* Removed `'mistral-v2'` option as no (open) models ever used it
* Removed all references to 'v2' template from comments
* Update llama.cpp
Fixed `trim_assistant_message` bug
* vulkan: Optimize soft_max
Large soft_max could already saturate memory, but small/medium sizes were
pretty slow. The bulk of the gains for them comes from using a smaller
workgroup size, and making the workgroup size match the subgroup size also
makes the barriers much cheaper.
Cache some values in locals to avoid refetching/recomputing. And stamp
out a few "template instantiations" so smaller cases will fully unroll.
Add a missing early return for OOB rows. This happens when there are more
than 512 rows and the dispatch is 512 x H.
* vulkan: Further soft_max optimizations
Restore the workgroup size of 512 case, use it for >1024.
Use unrollable loops for more iteration counts.
* tests: Fix memory bandwidth calculation for perf tests
Add a flops calculation for flash attention.
Add one GGML_OP_CPY perf test.
* vulkan: Optimize contiguous copies
Add a variant of the copy shader for when the tensors are contiguous. Avoid
the complex addressing calculations, and do four elements per invocation
to hide some other overhead.
Apply similar changes to the scale shader, since scale is always contiguous.
Add a "progress bar" for shader compiles.
* ggml : add ggml_flash_attn_ext_get_prec
* metal : use F16 precision in FA kernels
ggml-ci
* metal : minor clean-up
* metal : compile-guard bf16 FA kernels
ggml-ci
* build : remove obsolete compile flag [no ci]
* metal : prevent int overflows [no ci]
* cuda : disable BF16 FA
ggml-ci
* metal : fix BF16 requirement for FA kernels
ggml-ci
* make : clean-up [no ci]
* server : simple chat UI with vuejs and daisyui
* move old files to legacy folder
* embed deps into binary
* basic markdown support
* add conversation history, save to localStorage
* fix bg-base classes
* save theme preferences
* fix tests
* regenerate, edit, copy buttons
* small fixes
* docs: how to use legacy ui
* better error handling
* make CORS preflight more explicit
* add GET method for CORS
* fix tests
* clean up a bit
* better auto scroll
* small fixes
* use collapse-arrow
* fix closeAndSaveConfigDialog
* small fix
* remove console.log
* fix style for <pre> element
* lighter bubble color (less distract when reading)
* rwkv6: rename to wkv6
* rwkv6: support avx2 avx512 armv8 armv9
* rwkv6: update cuda file name
* rwkv6: rename params
* wkv on sycl
* sycl: add some ops
* sycl: Enhance OP support judgment
* wkv6: drop armv9 and tranfer to GGML style
ggml-ci
* sync : ggml
* update the function to use appropriate types
* fix define error
* Update ggml/src/ggml-cpu.c
* add appropriate asserts
* move element-wise functions outside
* put the declaration outside the loop
* rewrite to be more inline with the common pattern for distributing threads
* use recommended way GGML_TENSOR_LOCALS
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Co-authored-by: Diego Devesa <slarengh@gmail.com>
Co-authored-by: Plamen Minev <pacominev@gmail.com>
Co-authored-by: Yuri Khrustalev <ykhrustalev@users.noreply.github.com>
Co-authored-by: Meng, Hengyu <airdldl@163.com>
* ggml : add initial BF16 support
ggml-ci
* metal : add mul_mat_id BF16 support
ggml-ci
* metal : check for bfloat support on the Metal device
ggml-ci
* metal : better var names [no ci]
* metal : do not build bfloat kernels when not supported
ggml-ci
* metal : try to fix BF16 support check
ggml-ci
* metal : this should correctly check bfloat support
* Add granite template to llama.cpp
* Add granite template to test-chat-template.cpp
* Update src/llama.cpp
Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
* Update tests/test-chat-template.cpp
Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
* Added proper template and expected output
* Small change to \n
Small change to \n
* Add code space &
Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
* Fix spacing
* Apply suggestions from code review
* Update src/llama.cpp
---------
Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
* llama : deprecate softmax sampler + fix dist sampler
ggml-ci
* tests : replace macros with functions
ggml-ci
* sampling : change temperature sampler logic
For t <= 0.0f, keep the max logit intact and set the rest to -inf
* cont : no need for special "greedy" logic
top-k == 1 is the same
* tests : init prob correctly
* llama : handle temp <= 0.0 in the temp_ext sampler too
ggml-ci
* cont : avoid extra loop in temperature sampler for sub-zero temp
ggml-ci
Prior to this commit, using a JSON Schema containing a string
with `pattern` regular expression that uses top-level alternation
(e.g. `"pattern": "^A|B|C|D$"`) would result in invalid JSON
output from the constrained sampling grammar, because it
ended up creating a grammar rule like this for the string:
```
thing ::= "\"" "A" | "B" | "C" | "D" "\"" space
```
Note that this rule will only match a starting quote for the "A" case,
and will only match an ending quote for the "D" case,
so this rule will always produce invalid JSON when used for sampling
(that is, the JSON will always be lacking the starting quote,
the ending quote, or both).
This was fixed in a simple way by adding parentheses to the
generated rule (for all string pattern rules, to keep it simple),
such that the new generated rule looks like this (correct):
```
thing ::= "\"" ("A" | "B" | "C" | "D") "\"" space
```
* Initial XTC commit
Adds XTC sampler, not activated by default, but recommended settings by default.
* Cleanup
* Simplified chances calculation
To be more inline with the original implementation, chance is calculated once at the beginning.
* First fixes by comments
Still need to look into sorting
* Fixed trailing backspaces
* Fixed RNG to be reproduceable
Thanks to @slaren for directions
* Fixed forgotten header
* Moved `min_keep`
Moved from conditions to a simple check at the end.
* Fixed broken randomization
Thanks to @slaren for explanation
* Swapped sorting for a custom algorithm
Shifts tokens to remove the penalized ones, then puts the penalized at the back. Should make `min_keep` still viable.
* Algorithm rework
1. Scan token from top till the first non-penalizable
2. Remove the last captured token (the least probable above threshold)
3. Shift all tokens to override the remaining penalizable
4. Penalize and put them at the the bottom.
* Added XTC to `test-sampling`
* Simplified algorithm and more tests
* Updated info in common and args
* Merged back lost commits in common and arg
* Update dump info in common
* Fixed incorrect min_keep check
* Added XTC to README
* Renamed parameters, fixed info and defaults
* probability is at 0 by default, but XTC is included in sampling queue
* threshold higher than 0.5 switches XTC off
* Initial server support
* Added XTC to server UIs
* Fixed labels in old server UI
* Made algorithm safer and more readable
* Removed xtc_threshold_max
* Fixed arg after update
* Quick fixes by comments
* Simplified algorithm since threshold_max is removed
* Renamed random distribution
* Fixed tests and outdated README
* Small fixes
* ggml : do not use BLAS with types without to_float
* ggml : return pointer from ggml_internal_get_type_traits to avoid unnecessary copies
* ggml : rename ggml_internal_get_type_traits -> ggml_get_type_traits
it's not really internal if everybody uses it
* test-backend-ops : use flops for some performance tests
- parallelize tensor quantization
- use a different set of cases for performance and correctness tests
- run each test for at least one second
* threadpool: skip polling for unused threads
Currently all threads do N polling rounds even if only 1 thread is active (n_threads_cur == 1).
This commit adds a check to skip the polling for unused threads (ith >= n_threads_cur).
n_threads_cur is now an atomic_int to explicitly tell thread sanitizer that it is written
from one thread and read from other threads (not a race conditions).
* threadpool: further simplify and improve ggml_barrier
Avoid using strict memory order while polling, yet make sure that all threads go through
full memory barrier (memory fence) on ggml_barrier entrace and exit.
* threads: add simple barrier test
This test does lots of small, parallel matmul ops where the barriers in between dominate the overhead.
* threadpool: improve thread sync for new-graphs
Using the same tricks as ggml_barrier. All the polling is done with relaxed memory order
to keep it efficient, once the new graph is detected we do full fence using read-modify-write
with strict memory order.
* threadpool: improve abort handling
Do not use threadpool->ec (exit code) to decide whether to exit the compute loop.
threadpool->ec is not atomic which makes thread-sanitizer rightfully unhappy about it.
Instead introduce atomic threadpool->abort flag used for this. This is consistent with
how we handle threadpool->stop or pause.
While at it add an explicit atomic_load for n_threads_cur for consistency.
* test-barrier: release threadpool before releasing the context
fixes use-after-free detected by gcc thread-sanitizer on x86-64
for some reason llvm sanitizer is not detecting this issue.
* tests: add gradient checking to test-backend-ops
* remove old comment
* reorder includes
* adjust SIN/COS parameters
* add documentation, use supports_op if possible
* ggml_cont: fix issue with transposed tensors when one dimension is 1
when using multiple threads, it is not enough
to check for the tensors to be contiguous for
ggml_compute_forward_dup_same_cont to work correctly.
The tensors strides also need to match.
Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>
* Add ggml_cont tests
Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>
* Remove dead code
it isn't possible to reach this code because
all these functions are invoked by ggml_compute_forward_dup
if and only if src0->type != dst->type
Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>
* Make ggml_compute_forward_dup_same_cont work with contiguous tensors
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>
---------
Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
- Add `struct llama_sampler` and `struct llama_sampler_i`
- Add `llama_sampler_` API
- Add `llama_sampler_chain_` API for chaining multiple samplers
- Remove `LLAMA_API_INTERNAL`
- Add `llama_perf_` API and remove old `llama_print_timings` and `llama_reset_timings`
* ggml-quants : 1.625 bpw ternary packing for BitNet 1.58b
* ggml-quants : faster 1.625 bpw AVX2 vec_dot
Not using a lookup table anymore makes it match q4_0 speed.
* gguf-py : fix formatting
* llama : remove spaces on empty line
* ggml-quants : subtract 1 when back in epi8
This makes the 1.625 bpw type go faster than q4_0. Still not the fastest.
* ggml-quants : Q2_2 now faster than Q4_K on with AVX2
* ggml-quants : cleanup Q1_3 code formatting
* ggml-quants : ARM NEON vec_dot for q2_2 and q1_3
* ggml-quants : use ceiling division when quantizing q1_3
* convert-hf : simplify BitNet pre-quantization
This still results in the exact same tensor weights and scales,
but it reveals some weirdness in the current algorithm.
* convert-hf : allow converting the weird BitNet 1.3B
Its FFN size is 5460 which is not convenient.
The offending tensors are kept in F16,
which makes the final model 5.01 bpw.
* bitnet : replace 1.58b with b1.58, as in the paper
* ggml-quants : fix build failure on Windows
* ggml-quants : attempt to fix Arm 32-bit support
* ggml : add some informative comments in q1_3 vec_dot
* ggml : add TQ1_0 and TQ2_0 ternary quantization types
* ggml : even faster TQ2_0
* ggml : also faster TQ1_0
Same optimization as for TQ2_0 by offsetting the sum instead of the weights.
This makes TQ1_0 almost as fast as Q8_0 on AVX2.
* ggml : fix build issues in certain environments
* ggml : add NEON vec_dot implementation for TQ1_0 and TQ2_0
* ggml : avoid directly using vmlal_high_s8, for 32-bit ARM compat
The compiler seems smart enough to use the same instruction
even when using vget_high_s8 instead.
* ggml : remove q1_3 and q2_2
No more 1.625 bpw and 2.000 bpw,
now instead using 1.6875 bpw and 2.0625 bpw
with TQ1_0 and TQ2_0, respectively.
* llama : remove the separate scale tensors of BitNet b1.58
They won't be needed, since the remaining ternary quant types have
built-in scales.
* ggml-quants : rename fields of TQ1_0 and TQ2_0 structs for consistency
* ggml-quants : allow using vdotq_s32 in TQ2_0 vec_dot
Not yet tested on hardware which supports it,
might not work or might not even compile. But also it might.
It should make the performance better on recent ARM CPUs.
* ggml-quants : remove comment about possible format change of TQ2_0
Making it slightly more convenient for AVX512
but less convenient for everything else is not worth the trouble.
* gguf-py : Numpy (de)quantization for TQ1_0 and TQ2_0
* ggml-quants : use roundf instead of nearest_int for TQ1_0 and TQ2_0
This does not change anything for ternary models,
since their values should never end up being in halfway cases anyway.
* convert : allow direct conversion to TQ1_0 and TQ2_0
The token embeddings and output tensors are kept in F16
to allow quantizing them to Q4_K and Q6_K with llama-quantize.
* llama : handle fallback for TQ1_0 and TQ2_0 with Q4_0
Q4_0 is not completely symmetric (so not lossless for ternary models),
but it should be good enough.
* ggml-quants : allow using ARM dot product instructions for TQ1_0
* ggml-quants : deduplicate TQ1_0 and TQ2_0 __ARM_FEATURE_DOTPROD support
* ggml : remove unused ggml_mul special case
It would otherwise conflict with the more general
optimization coming with Mamba-2.
* ggml : handle TQ1_0 and TQ2_0 in dequantization-based operators
* test-backend-ops : add TQ1_0 and TQ2_0 comments for later
Not yet adding uncommented, because some backends like SYCL and Metal
do not properly handle unknown types in supports_op for GGML_OP_MUL_MAT.
(and Metal also doesn't handle it with GGML_OP_GET_ROWS)
Support for TQ1_0 and TQ2_0 for other backends than CPU
will be added in follow-up pull requests.
* Introduce ggml_compute_threadpool
- OpenMP functional: check
- Vanilla ggml functional: Check
- ggml w/threadpool functional: Check
- OpenMP no regression: No glaring problems
- Vanilla ggml no regression: No glaring problems
- ggml w/threadpool no regression: No glaring problems
* Minor fixes
* fixed use after release bug
* fixed a harmless race condition
* Fix Android bulid issue
* fix more race conditions
* fix deadlock for cases where cgraph.n_nodes == 1
and fix --poll case
* threadpool: use cpu_get_num_math to set the default number of threadpool threads
This way we avoid using E-Cores and Hyperthreaded siblings.
* bench: create fresh threadpool for each test
For benchmarking it's better to start a fresh pool for each test with the exact number of threads
needed for that test. Having larger pools is suboptimal (causes more load, etc).
* atomics: always use stdatomics with clang and use relaxed memory order when polling in ggml_barrier
This also removes sched_yield() calls from ggml_barrier() to match OpenMP behavior.
* threadpool: make polling the default to match openmp behavior
All command line args now allow for setting poll to 0 (false).
* threadpool: do not wakeup threads in already paused threadpool
* fix potential race condition in check_for_work
* threadpool: do not create two threadpools if their params are identical
* threadpool: reduce pause/resume/wakeup overhead in common cases
We now start threadpool in paused state only if we have two.
The resume is now implicit (ie new work) which allows for reduced locking and context-switch overhead.
* threadpool: add support for hybrid polling
poll params (--poll, ...) now specify "polling level", i.e. how aggresively we poll before waiting on cond.var.
poll=0 means no polling, 1 means poll for 128K rounds then wait, 2 for 256K rounds, ...
The default value of 50 (ie 50x128K rounds) seems like a decent default across modern platforms.
We can tune this further as things evolve.
* threadpool: reduce the number of barrier required
New work is now indicated with an atomic counter that is incremented for
each new graph that needs to be computed.
This removes the need for extra barrier for clearing the "new_work" and
removes the special case for trivial graphs.
* threadpool: remove special-casing for disposable threadpools
With the efficient hybrid polling there is no need to make disposable pools any different.
This simplifies the overall logic and reduces branching.
Include n_threads in debug print for disposable threadpool.
Declare pause and stop flags as atomic_bool
This doesn't actually generate any memory barriers and simply informs
the thread sanitizer that these flags can be written & read by different
threads without locking.
* threadpool: do not clear barrier counters between graphs computes (fixes race with small graphs)
This fixes the race condition with very small graphs where the main thread happens to
start a new graph while the workers are just about to exit from barriers.
* threadpool: use relaxed order for chunk sync
Full memory barrier is an overkill for this since each thread works on different chunk
* threadpool: remove abort_callback from threadpool state
* threadpool: better naming for thread/cpumask releated functions
* threadpool: consistent use of int type for n_threads params
* threadpool: add support for ggml_threadpool_params_default/init
Also removes the need for explicit mask_specified param.
all-zero cpumask means use default (usually inherited) cpu affinity mask.
* threadpool: move typedef into ggml.h
* threadpool: fix apply_priority() function name
* threadpool: fix swift wrapper errors due to n_threads int type cleanup
* threadpool: enable --cpu-mask and other threadpool related options only if threadpool is enabled
* threadpool: replace checks for compute_thread ret code with proper status check
* threadpool: simplify threadpool init logic and fix main thread affinity application
Most of the init code is now exactly the same between threadpool and openmp.
* threadpool: update threadpool resume/pause function names
* threadpool: enable openmp by default for now
* threadpool: don't forget to free workers state when omp is enabled
* threadpool: avoid updating process priority on the platforms that do not require it
On Windows we need to change overall process priority class in order to set thread priorities,
but on Linux, Mac, etc we do not need to touch the overall process settings.
* threadpool: update calling thread prio and affinity only at start/resume
This avoids extra syscalls for each graph_compute()
* llama-bench: turn threadpool params into vectors, add output headers, etc
* llama-bench: add support for cool off between tests --delay
This helps for long running tests on platforms that are thermally limited (phones, laptops, etc).
--delay (disabled by default) introduces the sleep for N seconds before starting each test.
* threadpool: move process priority setting into the apps (bench and cli)
This avoids changing the overall process priority on Windows for the apps
that use ggml/llama.cpp directy.
* threadpool: move all pause/resume logic into ggml
* threadpool: futher api cleanup and prep for future refactoring
All threadpool related functions and structs use ggml_threadpool prefix.
* threadpool: minor indent fixes
* threadpool: improve setprioty error message
* Update examples/llama-bench/llama-bench.cpp
Co-authored-by: slaren <slarengh@gmail.com>
* threadpool: fix indent in set_threadpool call
* use int32_t for n_thread type in public llama.cpp API
* threadpool: use _new and _free instead of _create and _release
* fix two more public APIs to use int32_t for n_threads
* build: set _GNU_SOURCE for Adroid
---------
Co-authored-by: Max Krasnyansky <quic_maxk@quicinc.com>
Co-authored-by: fmz <quic_fzaghlou@quic.com>
Co-authored-by: Max Krasnyansky <max.krasnyansky@gmail.com>
Co-authored-by: slaren <slarengh@gmail.com>
* [CANN] Add Ascend NPU backend
Ascend is a full-stack AI computing infrastructure for industry
applications and services based on Huawei Ascend processors and
software.
CANN (Compute Architecture of Neural Networks), developped by
Huawei, is a heterogeneous computing architecture for AI.
Co-authored-by: wangshuai09 <391746016@qq.com>
* delete trailing whitespaces
* Modify the code based on review comment
* Rename LLAMA_CANN to GGML_CANN
* Make ggml-common.h private
* add ggml_cann prefix for acl funcs
* Add logging for CANN backend
* Delete Trailing whitespace
---------
Co-authored-by: wangshuai09 <391746016@qq.com>
* llama : fix mpt and olmo pre-tokenizer
* llama : pre-tokenize non-special user-defined tokens first
* llama : fix detection of control-like user-defined tokens
* convert_hf : identify which user-defined tokens are control tokens
Only used in _set_vocab_gpt2() for now.
* convert_hf : identify more added control tokens for SPM tokenziers
This makes Gemma and Gemma-2 tokenize pretty much EVERYTHING correctly,
including HTML tags and consecutive spaces,
but it unfortunately requires model re-conversion.
There seems to be a weird behavior of the HF tokenizer for Gemma,
which prefers to use the 16-space token over more lengthy space tokens,
while using the SentencePiece tokenizer does not do this.
(the implementation in llama.cpp has the same behavior as SentencePiece)
* llama : fix wrong pre-tokenization of byte tokens
* llama : fix Viking pre-tokenizer regex
The order was previously wrong, which caused errors in some tests.
* llama : fix command-r detokenization
* convert_hf : reduce usages of the UNKNOWN token type
* llama : add UNKNOWN tokens in the special tokens cache
* convert_hf : reduce usages of UNKNOWN for InternLM2
This makes the changes from #8321 more consistent
with the other changes made here.
* test-tokenizer-random : reduce potential confilcts with #8379
* test-tokenizer-random : add a failing edge case for falcon
* conv transpose 1d passing test for 1d input and kernel
* working for different input and output channel counts, added test for variable stride
* initial draft appears to work with stride other than 1
* working with all old and new conv1d tests
* added a test for large tensors
* removed use cuda hardcoding
* restored test-conv-transpose.c
* removed unused arugments, and fixed bug where test failure would cause subsequent tests to fail
* fixed accumulator bug
* added test to test-backend-ops
* fixed mistake
* addressed review
* fixed includes
* removed blank lines
* style and warning fixes
* return failure when test fails
* fix supports_op
---------
Co-authored-by: slaren <slarengh@gmail.com>
* py : type-check all Python scripts with Pyright
* server-tests : use trailing slash in openai base_url
* server-tests : add more type annotations
* server-tests : strip "chat" from base_url in oai_chat_completions
* server-tests : model metadata is a dict
* ci : disable pip cache in type-check workflow
The cache is not shared between branches, and it's 250MB in size,
so it would become quite a big part of the 10GB cache limit of the repo.
* py : fix new type errors from master branch
* tests : fix test-tokenizer-random.py
Apparently, gcc applies optimisations even when pre-processing,
which confuses pycparser.
* ci : only show warnings and errors in python type-check
The "information" level otherwise has entries
from 'examples/pydantic_models_to_grammar.py',
which could be confusing for someone trying to figure out what failed,
considering that these messages can safely be ignored
even though they look like errors.
* add chatglm3-6b model support huggingface model:
https://hf-mirror.com/THUDM/chatglm3-6b
Signed-off-by: XingXing Qiao <qiaoxx@dingdao.com>
* remove .rotary_pos_emb.inv_freq and unuse code for chatglm3 model
Signed-off-by: XingXing Qiao <qiaoxx@dingdao.com>
* fix lint error
Signed-off-by: XingXing Qiao <qiaoxx@dingdao.com>
* optimize convert-hf-to-gguf.py for chatglm model
Signed-off-by: XingXing Qiao <qiaoxx@dingdao.com>
* support glm-4-9b-chat
Signed-off-by: XingXing Qiao <qiaoxx@dingdao.com>
* fix eos tokens to glm4
* remove unused log
* add preprocess to chatglm3 and chatglm4
* add eos_id_list to llama.cpp
* fix code style
* fix code style
* fix conflicts
* fix conflicts
* Revert "add eos_id_list to llama.cpp"
This reverts commit 3a4d5790bf.
* set <|endoftext|> as eos and <|user|> as eot
* fix chat template bug
* add comment to glm prefix and suffix
* fix conflicts and add rope_ratio & ChatGLMForConditionalGeneration
* fix chat template bug
* fix codestyle
* fix conflicts
* modified the general name of glm model
* fix conflicts
* remove prefix and suffix
* use normal glm4 chattempalte & use LLM_FFN_SWIGLU in phi3
* fix: resolve Flake8 errors in `convert-hf-to-gguf.py`
- Fix E302 by adding two blank lines before top-level function definitions
- Replace print statements to fix NP100
- Fix E303 by ensuring only one blank line between lines of code
* fix rope ratio to solve incorrect answers
* fix by comments
---------
Signed-off-by: XingXing Qiao <qiaoxx@dingdao.com>
Co-authored-by: XingXing Qiao <qiaoxx@dingdao.com>
Co-authored-by: Umpire2018 <138990495+Umpire2018@users.noreply.github.com>
* Add llama_detokenize():
- Update header files location
- UNKNOWN and CONTROL are 'special pieces'
- Remove space after UNKNOWN and CONTROL
- Refactor llama_token_to_piece()
- Add flag: clean_up_tokenization_spaces
- Symmetric params for llama_tokenize() and llama_detokenize()
* Update and fix tokenizer tests:
- Using llama_detokenize()
- Unexpected vocab type as test fail instead of error
- Useful when automating tests:
- If you don't know in advance the vocab type
- Differenciate other loading errors
- Skip unicode surrogaes and undefined
- Gracefully exit threads
- Using exit() is throwing random exceptions
- Clean old known problematic codepoints
- Minor: confusing hexadecimal codepoint
* Update bruteforce random tests
- Add detokenizer checks
- New generator: ascii_lr_strip
- New generator: apostrophe
- Add more vocabs files
- Detokenize special tokens.
- Replace errors with '\uFFFD' when detokenizing to 'utf-8'
- More edge cases
- Better detokenization results check
* Fix add_space_prefix, set false by default
* Better leading space removal
* Do not remove space when decoding special tokens
* Bugfix: custom regexs splits undefined unicode codepoints
* 'viking' detokenizer clean spaces
* json: default additionalProperty to true
* json: don't force additional props after normal properties!
* json: allow space after enum/const
* json: update pydantic example to set additionalProperties: false
* json: prevent additional props to redefine a typed prop
* port not_strings to python, add trailing space
* fix not_strings & port to js+py
* Update json-schema-to-grammar.cpp
* fix _not_strings for substring overlaps
* json: fix additionalProperties default, uncomment tests
* json: add integ. test case for additionalProperties
* json: nit: simplify condition
* reformat grammar integ tests w/ R"""()""" strings where there's escapes
* update # tokens in server test: consts can now have trailing space
* llama : return nullptr from llama_grammar_init
This commit updates llama_grammar_init to return nullptr instead of
throwing an exception.
The motivation for this is that this function is declared inside an
extern "C" block and is intended/may be used from C code which will not
be able to handle exceptions thrown, and results in undefined behavior.
On Windows and using MSVC the following warning is currently generated:
```console
C:\llama.cpp\llama.cpp(13998,1): warning C4297: 'llama_grammar_init':
function assumed not to throw an exception but does
C:\llama.cpp\llama.cpp(13998,1): message :
__declspec(nothrow), throw(), noexcept(true), or noexcept was specified
on the function
```
Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
* squash! llama : return nullptr from llama_grammar_init
Add checks for nullptr when calling llama_grammar_init.
Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
---------
Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
Co-authored-by: Clint Herron <hanclinto@gmail.com>
* Adding simple bare-bones test for end-to-end integration test for json validation against auto-generated JSON-schema grammars.
* Adding additional examples as documented in #7789 . Also adding the ability to automatically output improperly failing grammars to debug output files so they can more easily be examined in the gbnf-validator program.
* Uncommenting formerly commented tests so that they fail for others who are attempting to reproduce the bugs.
* Merging improved schema test methods added by @ochafik in #7797
* Adding #define to temporarily remove failing tests so that this PR can pass CI, but still be useful for other PRs that want to leverage the framework.
* Fixing nits from ochafik. Removing escape slashes, adding additional failing cases, fixing some other strings.
* Fixing grammar indentation to be consistent throughout file.
* cuda sqrt support
* enable cuda in pca
* fix comments in pca
* add test
* add sqrt to ggml_backend_cuda_supports_op
* fix test
* new line
* Use F32 sqrtf instead of F64 sqrt
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
---------
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* Add per token attributes enum
* Using phi-3 for testing 'rstrip'
* Using jina-v2 for testing 'lstrip'
* Brute force test for 'lstrip' and 'rstrip'
* Implement 'rstrip' and 'lstrip'
* Update phi-3 GGUF file (obsolete since 917dc8c)
* Replace llama_token_type with llama_token_attribs
* ggml : fix loongson compile warnings
ggml-ci
* Fix loongarch quantize test fail.
Fix unexpected error introduced during rebase code.
* tests : disable json test due to lack of python on the CI node
ggml-ci
---------
Co-authored-by: junchao-loongson <zhaojunchao@loongson.cn>
* Update random test: add_bos_token.
* Update random test: add WPM models for testing.
* Build vocab.special_tokens_cache using vocab token types.
* Fix and improve WPM preprocessing.
- Fix unicode edge case combinations.
- Split by whitspace in the same pass.
* Discard all tokens when no matching found.
* Fix phi3 template matching vs zephyr
* Add regression test for new phi3 chat template
* Implement review suggestions
* Fix phi3 jinja test templates & match by <|end|>
* Apply suggestion
Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
* Add all phi3 template variants in tests
* Remove unneeded message trimming
Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
* Fix tests to not expect trimmed messages
---------
Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
* cuda : fix rope pos data
ggml-ci
* ggml : drop mode & 1 == 1 support for ggml_rope
ggml-ci
* ggml : support freq_factors for f16 rope (CPU)
ggml-ci
* tests : add rope tests using frequency factors
ggml-ci
* add phi3 128k support in convert-hf-to-gguf
* add phi3 128k support in cuda
* address build warnings on llama.cpp
* adjust index value in cuda long rope freq factors
* add long rope support in ggml cpu backend
* make freq factors only depend on ctx size
* remove unused rope scaling type 'su' frin gguf converter
* fix flint warnings on convert-hf-to-gguf.py
* set to the short freq factor when context size is small than trained context size
* add one line of comments
* metal : support rope freq_factors
* ggml : update ggml_rope_ext API to support freq. factors
* backends : add dev messages to support rope freq. factors
* minor : style
* tests : update to use new rope API
* backends : fix pragma semicolons
* minor : cleanup
* llama : move rope factors from KV header to tensors
* llama : remove tmp assert
* cuda : fix compile warning
* convert : read/write n_head_kv
* llama : fix uninitialized tensors
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update brute force test: add_special
* Update brute force test: default values for add_bos_token and add_eos_token
* Enable rtrim when pre-inserting BOS
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Revert "server : fix test regexes"
* Update brute force test: special tokens
* Fix added tokens
- Try to read 'added_tokens.json'.
- Try to read 'tokenizer_config.json'.
- Try to read 'tokenizer.json'.
* Fix special tokens rtrim
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* server : fix test regexes
* Replace CODEPOINT_TYPE_* with codepoint_flags
* Update and bugfix brute force random test
* Deterministic brute force random test
* Unicode normalization NFD
* Get rid of BOM
* initial commit with CPU implementation of upscale to shape and test, cuda implementation next
* experimental commit to see if dst shape is correct
* test version
* test
* removed unnecessary params
* refactor
* fixed tests
* ggml : metal impl + cleanup + sycl dev warnings
* patched ggml_upscale cuda op to handle non-contiguous tensors, added test for non-contiguous behavior
* metal : fix upsacle op to support nb00 + style
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Add left recursion check: quit early instead of going into an infinite loop
* Remove custom enum, rename left recursion check and move to "grammar internal" section, add handling for edge case where a leftmost nonterminal may be empty
* Remove unnecessary declaration
* Introduce bfloat16 support
Many models on Hugging Face (e.g. Mistral, TinyLLaMA) use bfloat16 as
their canonical floating point format.
┌sign
│
│ ┌exponent
│ │
│ │ ┌mantissa
│ │ │
│┌──┴───┐┌─┴───┐
0b0000000000000000 brain16
This encoding has the same number of exponent bits as float32. That
makes conversion relatively straightforward, even in the absence of
hardware support. For example, converting brain16 to binary32 means
simply shifting 16 bits to the left.
┌sign
│
│ ┌exponent
│ │
│ │ ┌mantissa
│ │ │
│┌──┴───┐┌─┴───────────────────┐
0b00000000000000000000000000000000 IEEE binary32
The issue is that converting bf16 to fp16 can result in information
loss. Only 13% of bf16 numbers can be precisely represented in fp16
which in practice ends up being 99.71% of Mistral 7b v0.2's weights
however there is currently no way other than fp32 to get the others
┌sign
│
│ ┌exponent
│ │
│ │ ┌mantissa
│ │ │
│┌─┴─┐┌─┴──────┐
0b0000000000000000 IEEE binary16
This change fixes that, by adding a bf16 data type to GGML. Support
for CPU inference has been implemented along with optimizations for
the AVX2, AVX512, and AVX512BF16 ISAs. Perplexity on Mistral 7b 0.2
improves somewhere around -0.0024 to -0.0046 compared to using fp16
* Remove GGML code that's not needed
* Minimize the GGML API surface area for BF16
* Remove bf16 luts
* Make the GGML header look nicer
* Fix documentation
* Apply ggerganov's fixes for test-backend-ops
* Add BF16 code for new ggml_validate_row_data() function
* convert.py: add python logging instead of print()
* convert.py: verbose flag takes priority over dump flag log suppression
* convert.py: named instance logging
* convert.py: use explicit logger id string
* convert.py: convert extra print() to named logger
* convert.py: sys.stderr.write --> logger.error
* *.py: Convert all python scripts to use logging module
* requirements.txt: remove extra line
* flake8: update flake8 ignore and exclude to match ci settings
* gh-actions: add flake8-no-print to flake8 lint step
* pre-commit: add flake8-no-print to flake8 and also update pre-commit version
* convert-hf-to-gguf.py: print() to logger conversion
* *.py: logging basiconfig refactor to use conditional expression
* *.py: removed commented out logging
* fixup! *.py: logging basiconfig refactor to use conditional expression
* constant.py: logger.error then exit should be a raise exception instead
* *.py: Convert logger error and sys.exit() into a raise exception (for atypical error)
* gguf-convert-endian.py: refactor convert_byteorder() to use tqdm progressbar
* verify-checksum-model.py: This is the result of the program, it should be printed to stdout.
* compare-llama-bench.py: add blank line for readability during missing repo response
* reader.py: read_gguf_file() use print() over logging
* convert.py: warning goes to stderr and won't hurt the dump output
* gguf-dump.py: dump_metadata() should print to stdout
* convert-hf-to-gguf.py: print --> logger.debug or ValueError()
* verify-checksum-models.py: use print() for printing table
* *.py: refactor logging.basicConfig()
* gguf-py/gguf/*.py: use __name__ as logger name
Since they will be imported and not run directly.
* python-lint.yml: use .flake8 file instead
* constants.py: logger no longer required
* convert-hf-to-gguf.py: add additional logging
* convert-hf-to-gguf.py: print() --> logger
* *.py: fix flake8 warnings
* revert changes to convert-hf-to-gguf.py for get_name()
* convert-hf-to-gguf-update.py: use triple quoted f-string instead
* *.py: accidentally corrected the wrong line
* *.py: add compilade warning suggestions and style fixes
* ggml : add ggml_flash_attn_ext API
* ggml : fix GQA support in ggml_flash_attn_ext
* ggml : online attention (CPU)
* metal : initial implementation
* metal : f16 precision
* metal : reduce branches
* metal : specialize for head size
* wip : 8 rows per simd group
* wip : 4 rows per simd group
* wip : template for rows per warp
* metal : parallelize across KV size
* metal : parallel reduce across heads
* metal : efficient flash_attn_f16 implementation
* metal : avoid redundant loads of the attention
* metal : scale and mask in matrix form
* metal : fix comment
* llama : avoid ggml_cast, use F32 query
* metal : add parallel reduce version (disabled)
* metal : move output into local memory + optimize
- the result from each simdgroup now stays in the registers
- significantly reduced SRAM usage
- more efficient skipping of -INF blocks
- avoid simdgroup barrier in hot loop
- add comments
* metal : add tests, fix scaling, support C > 32
* metal : improve precision
* ggml : fix f16 mad
* metal : minor
* metal : support Q > 8
* tests : add ATTN tests
* metal : disable buffer allocation logs
* tests : more
* metal : faster inner loop for C == 32
* metal : fix array initialization
* tests : ifdef
* ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext
* ggml : fix ggml_soft_max mask requirement
* cuda : fix soft_max to use correct mask size
* cuda : add flash_attn kernel (wip)
* metal : optimize softmax for C > 32
* metal : optimize softmax
* tests : minor fix
* cuda : avoid zeroing fragments
* tests : update dims
* cuda : fix __hisinf() result check
* cuda : avoid warp_reduce for smax
* cuda : use int instead of int64_t
Noticeably improves performance (thanks to Johannes)
* cuda : make loops use the same loop values
Thanks Johannes again for the tip
* cuda : unroll some of the loops
* cuda : avoid __hisinf branches
* cuda : use half2 in softmax
* cuda : switch to 1 warp for bs > 16
* cuda : speed-up reduce part of the kernel
* cuda : unroll Q*K^T loop
* cuda : fix -INF block check
* cuda : simplify softmax
* cuda : fix matrix names
* cuda : minor
* llama : adapt to F16 KQ_pos
* llama : adapt new models to F16 KQ_mask
* ggml : fix F16 store (ARM NEON)
* llama : fix type of KQ_mask and KQ_pos
* ggml : fix CPU soft_max
* tests : add hs=256
* cuda : fix build
* metal : improve perf via smaller int registers
* cuda : adapt soft_max to F16 mask and pos
* CUDA: faster FlashAttention, kernel for bs == 1
* 16 cols for Phi-2
* no vec for hs, no hs==256 ncols==32 for Volta
* adjust kernel selection logic
* 4 warps, 256 stride for all D
* no ncols == 64
* Multiple parallel blocks for batch size 1
* fix compile warnings
* fix excessive KQ_b loads
* fix cmake build
* fix KV cache padding, NaN from INFINITY (#6438)
* llama : flash_attn cparam + fix defrag
* server: support flash_attn param
* server: bench: enable flash_attn param
* CUDA: refactor host code, dyn. par. blocks
* fix flash_attn_vec_f16 race condition
* flush softmax exp below threshold to 0
* store temp KQ in registers
* Calculate KQ as FP32 if KQV has GGML_PREC_F32
* Add __hgt2_mask implementation for CUDA 11
* fix KQ FP32 precision fpr parallel_blocks > 1
* llama-bench : add -fa,--flash-attn arg
* metal : add BS=1 kernel for flash attention (#6508)
* metal : add BS=1 kernel for flash attention (wip)
* metal : support more than 1 warps
* metal : opts
* metal : opt
* metal : switch to parallel reduce
* metal : reduce registers
* metal : simplify
* metal : initial FA vec kernel
* metal : use F32 attention accumulators
* batched-bench : add fattn arg
* llama : simplify llama_build_kv_store
ggml-ci
* llama : adapt build_olmo to changes
* ggml : fix arm fp16 store on windows
* metal : clean-up
* metal : clean-up kernel code
* metal : minor
* tests : remove benchmarks
ggml-ci
* ggml : fix avx512 const correctness
ggml-ci
* ggml : fix soft_max with bias on CPU
ggml-ci
* common : print --flash-attn in help
* ggml : fix num dimensions in ggml_flash_attn_ext
* llama : force disable flash attention for incompatible models
* ggml : ggml_soft_max support F16/F32 mask/pos
ggml-ci
* cuda : uint -> uint32_t
* cuda : "constexpr dim3" -> "const dim3"
ggml-ci
* cuda : try to fix __hgt2_mask
ggml-ci
* ggml : add TODO's for F16/F32 mask/pos support in other backends
* llama : replace bool need_kq_pos with use_alibi
* llama : prep ALiBi support for BERT models
ggml-ci
* llama : fix n_batch requirements
ggml-ci
* cont
* server : add help for --flash-attn arg
* llama : disable FA for AMD
* tests : remove TMP_ATTN_BENCH
ggml-ci
* llama : support save/load state with FA enabled
ggml-ci
* ci : add CUDA save-load-state tests
ggml-ci
* llama : llama_kv_cache_clear zeroes data + fix save-load seq
ggml-ci
* llama : fix copy-paste errors, add TODO
* llama : disallow incompatible states
* llama : update llama_state_get_size after v_trans field
* metal : remove tmp log
* llama : add static reminder for llama_state_get_size
* metal : fix max nsg
ggml-ci
* ci : fix arg order
ggml-ci
---------
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
* Cleaning up integration tests to share code between tests and make it simpler to add new tests.
* Add tests around quantifiers to ensure both matching and non-matching compliance.
* Add slightly more complex grammar with quantifiers to test references with quantifiers.
* Fixing build when C++17 is not present.
* Separating test calls to give more helpful stack traces on failure. Adding verbose messages to give visibility for what is being tested.
* Adding quotes around strings to explicitly show whitespace
* Removing trailing whitespace.
* Implementing suggestions from @ochafik -- grammars and test strings now print and flush before tests to aid in debugging segfaults and whatnot.
* Cleaning up forgotten symbols. Modifying simple test to use test harness. Added comments for more verbose descriptions of what each test is accomplishing.
* Unicode symbol modifications to hopefully make log easier to parse visually.
* ggml : group all experts in a single ggml_mul_mat_id
cuda : improve mmid row copy
* cuda : fix bin bcast with non-cont src0
* test-backend-ops : only run all mul mat tests for base types
* llama : disable moe offloading with SYCL
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* support qwen2moe
* fix-review
* metal : support unary ops for nelements % 4 != 0
* metal : require contiguousness for float4 unary kernels
* metal : require contiguousness for float4 unary kernels (cont)
* fix-review
* names : for brevity "SHARED_EXP" -> "SHEXP"
* llama : reuse build_moe_ffn()
* llama : add model type name
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Add chat template for command-r model series
* Fix indentation
* Add chat template test for command-r models and update the implementation to trim whitespaces
* Remove debug print
* Added integration tests for GBNF parser to validate correctness of parsing, as well as correctness of string matching. Intended for use to pin behavior while working on performance improvements.
* Fixing whitespace errors and cleaning error message alert to be clearer.
* Removing hacky include to llama.cpp from grammar integration test now that needed functions are available via internal API.
* Comment cleanup.
* Reorganizing tests for readability.
* Cleaning up debug message to make a bit more sense.
* Add openchat chat template
* Add chat template test for openchat
* Add chat template for vicuna
* Add chat template for orca-vicuna
* Add EOS for vicuna templates
* Combine vicuna chat templates
* Add tests for openchat and vicuna chat templates
* Add chat template for alpaca
* Add separate template name for vicuna-orca
* Remove alpaca, match deepseek with jinja output
* Regenerate chat template test with add_generation_prompt
* Separate deepseek bos from system message
* Match openchat template with jinja output
* Remove BOS token from templates, unprefix openchat
* ggml : update mul_mat_id to use the same tensor for all the experts
* update cuda
* minor
* update metal
* update test-backend-ops
* fix cuda
* Update ggml-metal.m
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* update convert.py
* update convert-hf-to-gguf.py
* update convert.py for mixtral hf models
* Update convert-hf-to-gguf.py
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* cuda : support non-pow-2 number of experts
* allow quantize to work for split and merged experts models in the same way
* cleanup + disable mmap automatically with split tensors models
* update imatrix
* test-backend-ops : test qwen argsort
* update grok model loading
* llama : add merged experts tensors to the grok tensor map
* minor
* gguf : bump version
* fix quantizing of merged experts
* convert-hf-to-gguf.py : update grok (untested)
* make linter happy
* cuda/argsort : use shared memory instead of pool memory
* convert : fix grok tensor names
* metal : add support for non-pow-2 argsort
* llama : more loader cleanup, better error checking
* cuda : fix warning
* llama : still use mmap for loading old models, but copy the data to a host buffer
* add review note
* llama : remove ffn tensor counting + add sanity check
ggml-ci
* convert : fix handling of n_experts == None
ggml-ci
* imatrix : fix ncall counters
* llama : produce error if imatrix size does not match
* quantize : terminate on errors + trace logs
ggml-ci
* metal : pad shared memory to 16 bytes
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* iq1_m: basics
* iq1_m: basics-2
* iq1_m: CUDA dequantize works
Very 1st shot I get PPL = 9.76 for LLaMA-v2-7B.
* iq1_m: separate shifts for each group of 8 in a block
We get
PPL(LLaMA-v2-7B ) = 9.2810
PPL(LLaMA-v2-13B) = 6.8105
Not bad, but slightly higher than
sqrt(PPL(IQ1_S) * PPL(IQ2_XXS))
which is the expected outcome given that IQ1_M is
halfway between IQ1_S and IQ2_XXS in terms of bpw.
From this, we would expect
PPL = 9.14 for LLaMA-v2-7B
PPL = 6.63 for LLaMA-v2-13B
* iq1_m: go to 3-bit scales
There is slight increase in PPL, but the 0.0625 bpw reduction
in size is totally worth it.
We now have
PPL(LLaMA-v2-7B ) = 9.4469 at 1.96 bpw
PPL(LLaMA-v2-13B) = 6.8717 at 1.93 bpw
PPL(LLaMA-v2-70B) = 4.8568 at 1.85 bpw
* iq1_m: scalar dot product
* iq1_m: AVX2 dot product
* iq1_m: very slightly faster AVX2 dot product
* iq1_m: ARM_NEON dot product
Works, but very slow (10.5 t/s)
* iq1_m: Metal - dequantize works, dot product does not
* iq1_m: Metal now works
About the same performance as iq1_s.
* iq1_m: minor
* iq1_m: checking pure iq1_m quantization
It is pretty bad: PPL(LLaMA-v2-7B) = 34 if we quantize output.weight
with Q4_K.
* iiq1_m: slightly faster ARM_NEON dot product
10.5 t/s -> 11.65 t/s
* iq1_m: faster ARM_NEON dot product
11.65 t/s -> 14.9 t/s
* iq1_m: another minor ARM_NEON dot product improvement
14.9 -> 15.0 t/s
* iq1_m: small PPL improvement via super-block scale adjustment
After quantizing block scales redo the super-block scale fit.
PPL(LLaMA-v2-7B ) = 9.3346
PPL(LLaMA-v2-13B) = 6.8419
PPL(LLaMA-v2-70B) = 4.8294
PPL(Mistral-7B ) = 8.1624
* iq1_m: adapt to CUDA refactoring
* iq1_m: remove unused variable
We have progressed to warnings being errors.
* iq1_m: add to backend-ops tests
* iq1_m: fix Windows ARM
* iq1_m: use common definition of iq1m_scale_t
* cuda: assert -> NO_DEVICE_CODE
* iq1_M: PR comments
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
* Try IQ4_NL with blocks of 64 - does not look good
* iq4_xs: go to super-blocks of 256 and 6-bit scales for blocks of 32
* iq4_xs: CUDA works - 133.2 t/s
* iq4_xs: AVX2 dot product
* iq4_xs: ARM_NEON dot product
* iq4_nl: Metal implementation
As usual, Metal / Apple Silicon don't like my quants.
* iq3_xs: minor fix
* iq4_xs: shrink by using IQ3_S for attn_k and attn_q
* iq4_xs: revert using IQ3_S for attn_k and attn_v
PPL vs size is good, but CPU performance suffers: on M2 Max
TG-128 drops to 21.7 t/s from 28.8, and on a Ryzen-7950X
to 14.5 t/s from 15.8 t/s. On CUDA we have 135 t/s when
using IQ3_S vs 133 t/s with pure IQ4_XS.
* Fix CI
* iq4_xs: Added forgotten check for 256 divisibility
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
* Adding IQ2_S and IQ2_M as a single cumulative commit
* Update examples/quantize/quantize.cpp
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* iq4_nl: squash commits for easier rebase
* Basics (quantize, dequantize)
* CUDA dequantize and dot product
* Slightly faster CUDA dot product (120 t/s)
* Switch to 6-bit scales
* Scalar dot product
* AVX2 dot product
* ARM_NEON dot product
* Works on metal, but still slow
* Slightly better Metal dot product
* Another small Metal improvement
* Metal dot product is getting there
* Faster CUDA dot product
* Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided
* Report the actual bpw
* Add _xs mix that is 4.05 bpw for non-MoE models
* Remove IQ4_XS for now, slightly adjust kvalues_iq4nl
* AVX2 dot product uses Q8_0 instead of Q8_K
* Add to test-backend-ops
* Minor fix
* Also use use Q5_K for attn_output in MoE models
* Fixes after merging latest master
* Switching to blocks of 32
* AVX2 for blocks of 32
* Scaler dot product for blocks of 32
* ARM_NEON dot product for blocks of 32
* Metal kernels for blocks of 32
* Slightly faster Metal kernels
* Resurrecting iq3_xs
After all the experimentation, nothing was better than this.
* Minor PPL improvement via a block scale fudge factor
* Minor improvement via 3 neighbours
* iq3_xs: working scalar and AVX2 dot products
* iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s)
* iq3_xs: working Metal implementation
* Adding IQ3_M - IQ3_XS mix with mostly Q4_K
* iiq3_xs: a 3.4375 bpw variant
* iq3_xs: make CUDA work for new version
* iq3_xs: make scalar and AVX2 work for new version
* iq3_s: make ARM_NEON work with new version
* iq3_xs: make new version work on metal
Performance is very similar to Q3_K_S
* iq3_xs: tiny Metal speed improvement
* iq3_xs: tiny Metal speed improvement
* Fix stupid warning
* Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS
* iq3_xs: rename to iq3_s
* iq3_s: make tests pass
* Move Q3_K_XS mix to 3.25 bpw
* Attempt to fix failing tests
* Another attempt to fix the Windows builds
* Attempt to fix ROCm
* ROCm again
* iq3_s: partial fix for QK_K = 64
* iq3_s: make it work on metal for QK_K = 64
Pleasent surprise: the coding was super-block size independent,
so all it took was to delete some QK_K == 256 guards.
* Will this fix ROCm?
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
* server: fallback to chatml
* add new chat template
* server: add AlphaMonarch to test chat template
* server: only check model template if there is no custom tmpl
* remove TODO
* iq4_nl: squash commits for easier rebase
* Basics (quantize, dequantize)
* CUDA dequantize and dot product
* Slightly faster CUDA dot product (120 t/s)
* Switch to 6-bit scales
* Scalar dot product
* AVX2 dot product
* ARM_NEON dot product
* Works on metal, but still slow
* Slightly better Metal dot product
* Another small Metal improvement
* Metal dot product is getting there
* Faster CUDA dot product
* Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided
* Report the actual bpw
* Add _xs mix that is 4.05 bpw for non-MoE models
* Remove IQ4_XS for now, slightly adjust kvalues_iq4nl
* AVX2 dot product uses Q8_0 instead of Q8_K
* Add to test-backend-ops
* Minor fix
* Also use use Q5_K for attn_output in MoE models
* Fixes after merging latest master
* Switching to blocks of 32
* AVX2 for blocks of 32
* Scaler dot product for blocks of 32
* ARM_NEON dot product for blocks of 32
* Metal kernels for blocks of 32
* Slightly faster Metal kernels
* iq4_nl: Fix after merging with master
* iq4_nl: another fix after merging with master
* Use IQ4_NL instead of Q4_K when using k-quants is not possible
* Fix typo that makes several tests fail
* It was the ggml_vdotq thing missed inside the brackets
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
* llama: add llama_chat_apply_template
* test-chat-template: remove dedundant vector
* chat_template: do not use std::string for buffer
* add clarification for llama_chat_apply_template
* llama_chat_apply_template: add zephyr template
* llama_chat_apply_template: correct docs
* llama_chat_apply_template: use term "chat" everywhere
* llama_chat_apply_template: change variable name to "tmpl"
* iq1_s: WIP basics
* iq1_s: CUDA is working
* iq1_s: scalar CPU dot product
* iq1_s: WIP AVX2 dot product - something is not right
* Fix tests
* Fix shadow warnings
* Fix after merge with latest master
* iq1_s: AVX2 finally works
* iq1_s: ARM_NEON dot product. Works, but not very fast
* iq1_s: better grid
* iq1_s: use IQ2_XXS for attn_output
At a cost of 0.04 extra bpw this gives a big improvement in PPL.
* iq1_s: Metal basics
Dequantize works, but not dot product
* iq1_s: Metal works, but quite slow
As usual, Apple Silicon does not like the code I write.
* iq1_s: Tests
* iq1_s: slightly faster dot product
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
* ggml : avoid recomputing alibi slopes (CPU)
* llama : reuse hparams.f_max_alibi_bias in all cases
ggml-ci
* ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal)
ggml-ci
* ggml : handle all SRCs (do not break on first null)
ggml-ci
* tests : do not use slope for large soft_max
accumulates too much error
ggml-ci
* ggml : alternative ALiBi without extra tensor
We compute the slopes in the kernel
ggml-ci
* cuda : add ALiBi support in ggml_soft_max_ext
ggml-ci
* ggml : deprecate ggml_alibi
* ggml : support multi-sequence ALiBi (Metal)
ggml-ci
* cuda : add multi-seq ALiBi + remote F16 soft_max
ggml-ci
* ggml : update deprecation message
* ggml : fix pos ptr when no ALiBi
ggml-ci
* cuda : fix performance (pow -> powf)
* cuda : precompute ALiBi constants
* metal : pre-compute ALiBi slopes
ggml-ci
* llama : init kq_pos only if needed
ggml-ci
* test-backend-ops : add null pos test to soft_max
test-backend-ops : replace soft_max tests
ggml-ci
---------
Co-authored-by: slaren <slarengh@gmail.com>
* Added numa options to allow finer grained control as well as plumbing for a new mirror mode that will require numa.h
* Reverted Makefile
* Fixed include
* Removed sched.h from ggml.h, moved ggml_get_numa_affinity into ggml.c, removed trailing whitespace and fixed up a few inconsistent variables
* removed trailing whitespace
* Added numa options to allow finer grained control as well as plumbing for a new mirror mode that will require numa.h
* Reverting Makefile
* Fixed a number of issues with the move from BOOL to ggml_numa_strategies. Added a note about mirror mode note being implemented yet
* Removing MIRROR_MODE code for this PR
* Removing last bit of MIRROR_MODE code for this PR
* Removing unneeded branch in server.cpp example and moving get_numa_affinity and making it static
* Fixed lingering init_llama_backend() bool calls in tests and examples
* Remote enum llama_numa_strategies
* Revert bad merge with dynatemp flags
* add missing enum ggml_numa_strategies declaration and revert sync problem with master
* add missing enum ggml_numa_strategies declaration
* fixed ggml_init_numa variable
* Update ggml.h
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
* Update READMEs with info about numa flags, change INTERLEAVE strategy name to DISTRIBUTE everywhere, implement the improved distribution strategy from @rankaiyx, fix a spelling mistake and un-merge some bad merges
* split numa init out from llama_backend_init and created llama_numa_init. Updated all code paths and samples
* Fix up some boolean vs enum comparisons
* Added #ifdefs for non-Linux OS that don't have cpu_set_t datatype
* Update ggml.h
Align enum values
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update ggml.c
Remove whitespace
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update ggml.c
align paremeters
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update examples/server/server.cpp
remove whitespace and align brace
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update common/common.cpp
Remove whitespace and align brace
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* unified ggml_numa_strategy enum and fixed text alignment in server.cpp example
* Update ggml.c
simplified return for platforms without NUMA support
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
* removed redundant else from cli argument processing of --numa
* whitespace
---------
Co-authored-by: root <root@nenya.lothlorien.ca>
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Co-authored-by: Jared Van Bortel <jared@nomic.ai>
Xuan-Son Nguyen
R0CKSTAR
Jeff Bolz
Sergei Vorobyov
Georgi Gerganov
Michał Moskal
Gaurav Garg
Molly Sophia
Olivier Chafik
cmdr2
cmdr2
Diego Devesa
William Tambellini
Johannes Gäßler
Rémy O
Vinesh Janarthanan
bandoti
piDack
Akarshan Biswas
Xuan Son Nguyen
Yun Dou
Valentin Mamedov
HimariO
PAB
ltoniazzi
Juk Armstrong
Zhiyuan Li
arch-btw
wwoodsTM
Jun Hee Yoo
Joe Eli McIlvain
MaggotHATE
Zhenwei Jin
Max Krasnyansky
Salvatore Mesoraca
compilade
Faisal Zaghloul
zhentaoyu
fairydreaming
0cc4m
Mengqing Cao
hipudding
John Balis
toyer
jaime-m-p
Clint Herron
Daniel Bevenius
Calvin Laurenson
Tristan Druyen
liuwei-git
Haggai Nuchi
Haoxiang Fei
Ren Xuancheng
Justine Tunney
DAN™
Brian
Wouter
Shijie
Chao Jiang
kaizau
Kawrakow
leejet
Herman Semenov
bmwl