ggerganov/llama.cpp

Pull Requests Commits

cont : add TODO in clip.cpp

ggerganov committed 2 hours ago

1f6da4e7

llama: request GGML_PREC_BF16 for the KQ multiplication

ServeurpersoCom committed 2 hours ago

857cf9fc

tests: cover GGML_PREC_BF16 in the flash attention tests

ServeurpersoCom committed 2 hours ago

389d4a51

ggml: handle GGML_PREC_BF16 the same way as GGML_PREC_F32 in backends

ServeurpersoCom committed 3 hours ago

32c54f2d

llama, mtmd: follow the reference precision of ggml_prec_set_acc calls

ServeurpersoCom committed 3 hours ago

8c91606e

webgpu : Implement group norm kernel (#31490)

nikhilJain17 committed 5 hours ago

Verified cf659f24

vulkan: Add 'readonly' to decodeBuf declarations. (#31833)

jeffbolznv committed 5 hours ago

Verified 19381526

CUDA: fix shared memory bank conflicts in the lightning indexer (#31803)

aalramadan committed 5 hours ago

Verified 5d7d4aec

llama : add missing link to pthread (#31785)

angt committed 6 hours ago

Verified c0ab6e5c

opencl: add arange (#31829)

shaofeiqi committed 7 hours ago

Verified 31424b7c

meta : raise the compute view limit for buffers with few tensors (#31447)

developer0hye committed 9 hours ago

Verified fd27b203

hexagon: fuse post-norm residual add into RMS_NORM_MUL (#31822)

ebateni committed 12 hours ago

Verified ca89968a

model: add Kimi-K3.1-Pro text model (#31779)

usrlocalben committed 14 hours ago

Verified bcb01528

hexagon: conv_2d optimizations and support for conv_3d (#31814)

marktodorovich committed 17 hours ago

Verified 669204ee

docker: add explicit /usr/local/nvidia search path (#31806)

ngxson committed 19 hours ago

Verified 954de248

cli : fix crash on invalid UTF-8 in prompt or files (#31563)

SofianElmotiem committed 1 day ago

Verified 6c01da62

tests : qualify isnan and signbit with std:: (#31791)

ggerganov committed 1 day ago

Verified 8820b088

tests : override hardware_concurrency in CI (#31797)

ggerganov committed 1 day ago

Verified 9a1d03c7

cmake : generate RC file directly without win-app.rc.in (#31718)

angt committed 1 day ago

Verified 18ab62a8

opencl: support repeat for f16, i32, i16, bf16 (#31782)

lhez committed 1 day ago

Verified 52c9bad9

webgpu: add batched tensor API (`get_tensor_n` and `set_tensor_n`) (#31473)

yomaytk committed 1 day ago

Verified 5d3ed2ae

vulkan: detect ReBAR support, avoid BAR allocations if unavailable (#31658)

0cc4m committed 1 day ago

Verified 80088fa6

moe-cache : LRFU/MQ-like hybrid cache, ClassLRU (#31233)

sanmai committed 1 day ago

Verified c4fb5796

ci : bump CUDA 13 release builds to 13.5 (#31788)

sam-india-007 committed 1 day ago

Verified 3ed7284f

vendor: update BoringSSL to 0.20261105.0 (#31757)

cabelo committed 1 day ago

Verified d9cf1870

CUDA: fix Maxwell build (#31787)

JohannesGaessler committed 1 day ago

Verified 240d68fc

common : restrict HF token to huggingface.co hosts (#31730)

angt committed 1 day ago

Verified bbc2e4b6

opencl: add q6_k multi column tiled GEMV for speculative verification and refactoring (#31758)

shawngu-quic committed 1 day ago

Verified 4b77dee0

hexagon: Implement set_tensor_n and improve repacking/tiling robustness (#31744)

max-krasnyansky committed 1 day ago

Verified 67107ac3

mtmd: enable cmake unity build (#31760)

ngxson committed 1 day ago

Verified e14a7797

Older