ggml-org/llama.cpp

Pull Requests Commits

disable fwht shader on Intel Windows due to driver bug

0cc4m committed 6 hours ago

e6d79f12

disable subgroup shuffle on MoltenVK AMD

0cc4m committed 22 hours ago

6833f867

don't use N as workgroup size

0cc4m committed 23 hours ago

479618c9

vulkan: add fwht support for Intel with shmem reduction

0cc4m committed 1 day ago

3beb73cd

ci : add missing Linux label to cpu-x64-high-perf runner (#23958)

ggerganov committed 1 day ago

Verified e22b0de6

[SYCL] Support Q4_1, Q5_0, Q5_1 in Flash-attention (#23812)

arthw committed 1 day ago

Verified a5114249

[SYCL] Add more types in GET_ROWS OP (#23710)

arthw committed 1 day ago

Verified 41625226

sycl : Optimize Q3_K mul_mat by reorder (#23725)

arthw committed 1 day ago

Verified 44e211ce

ci: remove redundant or duplicate jobs (#23927)

netrunnereve committed 1 day ago

Verified af6528e6

server : handle If-None-Match weak ETags (#23916)

EZForever committed 1 day ago

Verified 6f165c1c

ci : limit trigger paths for the CPU workflow (#23938)

ggerganov committed 1 day ago

Verified 399739d5

vocab : add tokenizer support for jina-embeddings-v2-base-zh (#18756)

o7si committed 2 days ago

Verified d4c8e2c2

ui: fix ETag truncation with MSVC compiler (#23917)

EZForever committed 2 days ago

Verified 3292da09

docs : update ZenDNN docs for Q8 support (#23791)

truecoder34 committed 2 days ago

Verified e6123e20

llama: only use one iGPU device by default (#23897)

0cc4m committed 2 days ago

Verified 22cadc19

webui: add custom CSS injection via config (#23904)

ServeurpersoCom committed 2 days ago

Verified d749821d

Support `-fa auto` in llama-bench (#23714)

gaugarg-nv committed 2 days ago

Verified aa46bda8

opencl: support bf16 by converting to f16 (#23839)

lhez committed 2 days ago

Verified d6588daa

ui: exclude generated build dirs from prettier and eslint so lint errors stop being masked (#23910)

ServeurpersoCom committed 2 days ago

Verified d38d50e7

TP: fix granularity for Qwen 3.5/3.6 + 3 GPUs (#23843)

JohannesGaessler committed 2 days ago

Verified 8b0e0db6

metal : restore im2col implementation for large kernels (#23901)

ggerganov committed 3 days ago

Verified 2d9b7c8e

test: (test-llama-archs) log the config name first (#23885)

ngxson committed 3 days ago

Verified e674b127

ci : update ios-xcode release job to macos-26 (#23906)

ggerganov committed 3 days ago

Verified 4c4e91b7

ggml : add some lsx support (#23798)

MQ-mengqing committed 3 days ago

Verified d48a56ef

vulkan: add Flash Attention support for BFloat16 KV cache (#23420)

0cc4m committed 3 days ago

Verified 6e093b80

ci : fix s390x release job (#23898)

ggerganov committed 3 days ago

Verified 33752857

ci : clear cache instead of "no timestamp" keys + fix macos (#23895)

ggerganov committed 3 days ago

Verified d4204b03

llama : do not skip iGPU when only RPC devices are present (#23868)

rgerganov committed 3 days ago

Verified 1738129b

server: in SSE mode, send HTTP headers when slot starts (#23884)

ngxson committed 3 days ago

Verified 0821c5fc

ggml-webgpu: Check earlier for WebGPU required features (#23879)

reeselevine committed 3 days ago

Verified 151f3a98

Older