HappyZ happyz
0 Followers · 0 Following
Repositories
18
 Projects Packages Public Activity Starred Repositories
happyz synced commits to refs/pull/20394/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:27 -07:00
5e2ac9b851 Merge 07a6fd8775 into 6b949d1078
6b949d1078 sycl : support nvfp4 type in mul_mat (#21227)
84f82e846c ggml-cuda: Add generic NVFP4 MMQ kernel (#21074)
e1cb817483 memory: respect unified KV cache in hybrid memory for eval tasks (#21224)
88d5f8ffc3 CUDA/HIP: Fix kernel slection for mmvq mmid kernel to align host selection with device launch bounds (#21238)
Compare 10 commits »
happyz synced commits to refs/pull/20454/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:27 -07:00
ed1775ad32 Merge 97d973a8e5 into 6b949d1078
6b949d1078 sycl : support nvfp4 type in mul_mat (#21227)
84f82e846c ggml-cuda: Add generic NVFP4 MMQ kernel (#21074)
e1cb817483 memory: respect unified KV cache in hybrid memory for eval tasks (#21224)
88d5f8ffc3 CUDA/HIP: Fix kernel slection for mmvq mmid kernel to align host selection with device launch bounds (#21238)
Compare 10 commits »
happyz synced commits to refs/pull/20456/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:27 -07:00
de0229b918 Merge cf7b7d2f78 into 6b949d1078
6b949d1078 sycl : support nvfp4 type in mul_mat (#21227)
84f82e846c ggml-cuda: Add generic NVFP4 MMQ kernel (#21074)
e1cb817483 memory: respect unified KV cache in hybrid memory for eval tasks (#21224)
88d5f8ffc3 CUDA/HIP: Fix kernel slection for mmvq mmid kernel to align host selection with device launch bounds (#21238)
Compare 11 commits »
happyz synced commits to refs/pull/20487/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:27 -07:00
45e3fe1154 Merge df10279ece into 825eb91a66
825eb91a66 ggml-webgpu: port all AOT operators to JIT (#20728)
0fcb3760b2 fix: Use lower-case proxy headers naming (#21235)
6307ec07d3 common : cleanup logs and modernize the progress bar (#21215)
632219af73 CANN: fix multi-thread set_tensor race conditions (#20151)
Compare 35 commits »
happyz synced commits to refs/pull/20275/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:26 -07:00
f4eebe8ba0 Merge 981c80f2cd into 6b949d1078
6b949d1078 sycl : support nvfp4 type in mul_mat (#21227)
84f82e846c ggml-cuda: Add generic NVFP4 MMQ kernel (#21074)
e1cb817483 memory: respect unified KV cache in hybrid memory for eval tasks (#21224)
88d5f8ffc3 CUDA/HIP: Fix kernel slection for mmvq mmid kernel to align host selection with device launch bounds (#21238)
Compare 10 commits »
happyz synced commits to refs/pull/20269/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:26 -07:00
75225a93d6 Merge cfd5d05816 into 825eb91a66
825eb91a66 ggml-webgpu: port all AOT operators to JIT (#20728)
0fcb3760b2 fix: Use lower-case proxy headers naming (#21235)
6307ec07d3 common : cleanup logs and modernize the progress bar (#21215)
632219af73 CANN: fix multi-thread set_tensor race conditions (#20151)
Compare 18 commits »
happyz synced commits to refs/pull/20242/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:25 -07:00
0f636e9b4f Merge 205cda5ca6 into 88d5f8ffc3
88d5f8ffc3 CUDA/HIP: Fix kernel slection for mmvq mmid kernel to align host selection with device launch bounds (#21238)
d43375ff7f ggml : fix RWKV ops thread assignment (#21226)
2b86e5cae6 ggml-cpu: fix fallback for RVV kernels without zvfh (#21157)
88458164c7 CUDA: Add Flash Attention Support for Head Dimension 512 (#20998)
Compare 41 commits »
happyz synced commits to refs/pull/20075/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:24 -07:00
33077538e8 Merge 8a6b1c860c into 6b949d1078
6b949d1078 sycl : support nvfp4 type in mul_mat (#21227)
84f82e846c ggml-cuda: Add generic NVFP4 MMQ kernel (#21074)
e1cb817483 memory: respect unified KV cache in hybrid memory for eval tasks (#21224)
88d5f8ffc3 CUDA/HIP: Fix kernel slection for mmvq mmid kernel to align host selection with device launch bounds (#21238)
Compare 10 commits »
happyz synced commits to refs/pull/20238/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:24 -07:00
ad7f3ca3ab Merge 14a2ca7a21 into 6b949d1078
6b949d1078 sycl : support nvfp4 type in mul_mat (#21227)
84f82e846c ggml-cuda: Add generic NVFP4 MMQ kernel (#21074)
e1cb817483 memory: respect unified KV cache in hybrid memory for eval tasks (#21224)
88d5f8ffc3 CUDA/HIP: Fix kernel slection for mmvq mmid kernel to align host selection with device launch bounds (#21238)
Compare 10 commits »
happyz synced commits to refs/pull/20086/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:24 -07:00
dcf1342341 Merge c3fb035933 into d43375ff7f
d43375ff7f ggml : fix RWKV ops thread assignment (#21226)
2b86e5cae6 ggml-cpu: fix fallback for RVV kernels without zvfh (#21157)
88458164c7 CUDA: Add Flash Attention Support for Head Dimension 512 (#20998)
4951250235 llama : refactor llama_model_quantize_params to expose a pure C interface (#20346)
Compare 7 commits »
happyz synced commits to refs/pull/20112/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:24 -07:00
88da3136c0 Merge 2cf3eaf094 into 4951250235
4951250235 llama : refactor llama_model_quantize_params to expose a pure C interface (#20346)
82764c341a ggml webgpu: quantized buffers to u32 + wider browser/device support (#21046)
825eb91a66 ggml-webgpu: port all AOT operators to JIT (#20728)
0fcb3760b2 fix: Use lower-case proxy headers naming (#21235)
Compare 5 commits »
happyz synced commits to refs/pull/20064/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:23 -07:00
53b5350a07 Merge c3e73990ec into 825eb91a66
825eb91a66 ggml-webgpu: port all AOT operators to JIT (#20728)
0fcb3760b2 fix: Use lower-case proxy headers naming (#21235)
6307ec07d3 common : cleanup logs and modernize the progress bar (#21215)
632219af73 CANN: fix multi-thread set_tensor race conditions (#20151)
Compare 35 commits »
happyz synced commits to refs/pull/19855/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:23 -07:00
480308dd10 Merge e2df42ddfc into 6b949d1078
6b949d1078 sycl : support nvfp4 type in mul_mat (#21227)
84f82e846c ggml-cuda: Add generic NVFP4 MMQ kernel (#21074)
e1cb817483 memory: respect unified KV cache in hybrid memory for eval tasks (#21224)
88d5f8ffc3 CUDA/HIP: Fix kernel slection for mmvq mmid kernel to align host selection with device launch bounds (#21238)
Compare 10 commits »
happyz synced commits to refs/pull/19938/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:23 -07:00
5ee151d7f4 Merge 1842356d82 into 4951250235
4951250235 llama : refactor llama_model_quantize_params to expose a pure C interface (#20346)
82764c341a ggml webgpu: quantized buffers to u32 + wider browser/device support (#21046)
825eb91a66 ggml-webgpu: port all AOT operators to JIT (#20728)
0fcb3760b2 fix: Use lower-case proxy headers naming (#21235)
Compare 31 commits »
happyz synced commits to refs/pull/20009/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:23 -07:00
a3b5e4ef42 Merge b47e0cf21a into 825eb91a66
825eb91a66 ggml-webgpu: port all AOT operators to JIT (#20728)
0fcb3760b2 fix: Use lower-case proxy headers naming (#21235)
6307ec07d3 common : cleanup logs and modernize the progress bar (#21215)
632219af73 CANN: fix multi-thread set_tensor race conditions (#20151)
Compare 14 commits »
happyz synced commits to refs/pull/20062/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:23 -07:00
34ee95da28 Merge 6403785b94 into 6b949d1078
6b949d1078 sycl : support nvfp4 type in mul_mat (#21227)
84f82e846c ggml-cuda: Add generic NVFP4 MMQ kernel (#21074)
e1cb817483 memory: respect unified KV cache in hybrid memory for eval tasks (#21224)
88d5f8ffc3 CUDA/HIP: Fix kernel slection for mmvq mmid kernel to align host selection with device launch bounds (#21238)
Compare 10 commits »
happyz synced commits to refs/pull/19671/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:22 -07:00
d51abadc8b Merge 4e3eb50b7b into 84f82e846c
84f82e846c ggml-cuda: Add generic NVFP4 MMQ kernel (#21074)
e1cb817483 memory: respect unified KV cache in hybrid memory for eval tasks (#21224)
88d5f8ffc3 CUDA/HIP: Fix kernel slection for mmvq mmid kernel to align host selection with device launch bounds (#21238)
d43375ff7f ggml : fix RWKV ops thread assignment (#21226)
Compare 38 commits »
happyz synced commits to refs/pull/19763/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:22 -07:00
43174bea3d Merge 4d7c07a2f3 into 825eb91a66
825eb91a66 ggml-webgpu: port all AOT operators to JIT (#20728)
0fcb3760b2 fix: Use lower-case proxy headers naming (#21235)
6307ec07d3 common : cleanup logs and modernize the progress bar (#21215)
632219af73 CANN: fix multi-thread set_tensor race conditions (#20151)
Compare 23 commits »
happyz synced commits to refs/pull/19743/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:22 -07:00
b0670c6a16 Merge ef80e915e5 into 4951250235
4951250235 llama : refactor llama_model_quantize_params to expose a pure C interface (#20346)
82764c341a ggml webgpu: quantized buffers to u32 + wider browser/device support (#21046)
825eb91a66 ggml-webgpu: port all AOT operators to JIT (#20728)
0fcb3760b2 fix: Use lower-case proxy headers naming (#21235)
Compare 5 commits »
happyz synced commits to refs/pull/19755/merge at happyz/llama.cpp from mirror 2026-04-01 07:02:22 -07:00
cc689f9042 Merge 515bd7c9a5 into 88458164c7
88458164c7 CUDA: Add Flash Attention Support for Head Dimension 512 (#20998)
4951250235 llama : refactor llama_model_quantize_params to expose a pure C interface (#20346)
82764c341a ggml webgpu: quantized buffers to u32 + wider browser/device support (#21046)
825eb91a66 ggml-webgpu: port all AOT operators to JIT (#20728)
Compare 19 commits »