ggerganov/llama.cpp

Pull Requests Commits

CUDA: pick MMQ tile size against ncols_opt set on the host side

pwilkin committed 20 hours ago

f01a7549

CUDA: size routed MoE MMQ N-tiles from typical expert width on RDNA3 (#24546)

ravel7524 committed 20 hours ago

79800853

Revert "CUDA: size routed MoE MMQ N-tiles from typical expert width on RDNA3 (#24546)" (#28551)

pwilkin committed 20 hours ago

Verified e71b8051

nix : update deprecated expressions (#28145)

ZhaolunYin committed 21 hours ago

Verified ccc3646c

webgpu: format the GET_ROWS case block (#28542)

ServeurpersoCom committed 21 hours ago

Verified c0b1871b

server: fix LRU hang on multiple requests same model (#28539)

ngxson committed 21 hours ago

Verified 160bd031

sycl: add a batched L2_NORM kernel (#28222)

Titaniumtown committed 22 hours ago

Verified dbeb3754

vulkan: add DeepSeek-V4 hyper-connection fused ops (DSV4_HC_COMB/PRE/POST) (#26578)

kh0pper committed 22 hours ago

Verified 7a333e72

CUDA: size routed MoE MMQ N-tiles from typical expert width on RDNA3 (#24546)

ravel7524 committed 22 hours ago

Verified 0c963452

ggml: add gfx90c HIP support (#26454)

AuroraRAS committed 22 hours ago

Verified 47359973

convert : refactor Hy4-preview conversion - move HC tensor mapping to the global map (#28451)

fairydreaming committed 22 hours ago

Verified d23c47f2

CUDA: branchless Q4_K/Q5_K unpack to speed up mmvq, L2 prefetch on DGX Spark (#26705)

praneshgo committed 23 hours ago

Verified 73ab7599

vulkan: support type-aligned GET_ROWS (#28253)

jeffbolznv committed 1 day ago

Verified 0cae4306

examples : print ggml_version and ggml_commit in test-cmake [no ci] (#28538)

danbev committed 1 day ago

Verified 1173700b

caps : recheck typed content if template checks for string (#28511)

CISC committed 1 day ago

Verified 5202104b

convert : write explicit recurrent_layers for Qwen3-Next / Qwen3.5 (#28208)

devvexus committed 1 day ago

Verified 9a757058

ggml-cuda: fix divergent barrier in f16 flash attention (#27870)

siavashnorouzi committed 1 day ago

Verified b74f590e

ggml: allow backend inputs to not create another split (#28387)

am17an committed 1 day ago

Verified 992cb503

vulkan: rms_norm fusion opportunities (#28024)

jeffbolznv committed 1 day ago

Verified 9ac8c408

ci : add container image checking and tagging (wip) (#28394)

danbev committed 1 day ago

Verified 2092353c

vulkan: add TQ1_0 support (mm, mat-vec, mat-vec-id, dequant, get_rows) (#27765)

Anjielon committed 1 day ago

Verified 8fe90e1f

convert : add `--fuse-qkv` flag to fuse Q/K/V into QKV during HF-to-GGUF conversion (#22780)

JoursBleu committed 1 day ago

Verified 465e49b9

models : fix GDN normalization from `max` to `rsqrt` (#28068)

danielhanchen committed 1 day ago

Verified 5fdfa628

[Model] Support for Spark2_5ForCausalLM implementation (#27868)

KnightYao committed 1 day ago

Verified 3ad1ba73

opencl: properly choose weights pack for q4_K, q5_K mul_mat (#28402)

lhez committed 1 day ago

Verified d03efa5d

cuda: fixes races in mmid and mmf (#28475)

am17an committed 1 day ago

Verified 73a43d1f

grammar : fix max repetition threshold (#28469)

aldehir committed 2 days ago

Verified 9e0e2205

ui: Improve Chat Messages rendering performance (#28460)

allozaur committed 2 days ago

Verified 0afb805b

common: add --log-jsonl (#28437)

ngxson committed 2 days ago

Verified 7620399f

ui : embed assets directly with CMake (#28445)

angt committed 2 days ago

Verified c457e3bf

Older