ggerganov/llama.cpp

Pull Requests Commits

vulkan: when using transfer queue for async copies, sync on event_wait to avoid race

0cc4m committed 1 day ago

940d3276

hexagon: flash attention rework (optimizations, accuracy improvements, etc) (#25085)

max-krasnyansky committed 2 days ago

Verified 13e67386

CUDA: consistent use of __restrict__ + PDL for FA (#25185)

JohannesGaessler committed 2 days ago

Verified b820cc8e

ggml-cpu: add AVX2 optimization for nvfp4 dot product and use UE4M3 LUT (#23961)

ragz4125 committed 2 days ago

Verified 6dbc1174

ui Prevent tool messages from incorrectly appending to other conversations (#25177)

allozaur committed 2 days ago

Verified 9d88e7ce

ui: Remove PWA navigate fallback to prevent caching API endpoint requests (#25174)

allozaur committed 2 days ago

Verified 7af4279f

opencl: initial q1_0 support (#25160)

lhez committed 2 days ago

Verified fd1a0579

cuda : prevent integer truncation and overflow errors when using KQ mask strides in flash_attn_mask_to_KV_max kernel (#24945)

fairydreaming committed 2 days ago

Verified 0eca4d49

model : register t_layer_inp for qwen3next (#25141)

jschmied committed 3 days ago

Verified 4f31eedb

common,server: handle bracketed IPv6 literals in URL authority (#25140)

ServeurpersoCom committed 3 days ago

Verified 799fcc04

CUDA: fix get_rows_back for tables with more than 65535 rows (grid-y clamp + stride) (#25103)

mattjallo committed 3 days ago

Verified 931eb37f

CUDA: fix Gemma E4B MTP FlashAttention (#25148)

JohannesGaessler committed 3 days ago

Verified e495d1e7

vulkan: roll bk loop in matmul for asahi linux (#24663)

xingjianll committed 3 days ago

Verified f708a5b2

HIP: use hipBLAS for dense prefill on gfx900, keep MMQ for MoE (#24588)

DEV-DUFORD committed 3 days ago

Verified d9df1100

ggml-webgpu: add support for NVFP4 (#25143)

yomaytk committed 3 days ago

Verified 6c5de1cc

Revert "sched : reintroduce less synchronizations during split compute (#20793)" (#25138)

ORippler committed 3 days ago

Verified 86b94708

common : dedup preset and cached model entries in /v1/models (#25131)

angt committed 4 days ago

Verified 6f4f53f2

vulkan: use flops instead of weight tensor size for submission heuristic (#25005)

0cc4m committed 4 days ago

Verified 25a1d63f

DeepSeek V4 (#24162)

am17an committed 4 days ago

Verified 8c146a83

tools/ui: restore Tailwind scanning in ignored worktrees (#24879)

seryogakovalyov committed 4 days ago

Verified 6cb18b2f

common : remove unused regex-partial (#25118)

o7si committed 4 days ago

Verified 277a105d

jinja, chat: add --reasoning-preserve flag (#25105)

ngxson committed 4 days ago

Verified b3fed31b

Revert "ui: fix accessibility for hover-gated interactive elements assisted by claude(in debugging and tests) (#24727)" (#25098)

allozaur committed 4 days ago

Verified dbdaece2

ui: fix stop and reasoning skip in single-model mode (#25084)

ServeurpersoCom committed 4 days ago

Verified 7cb8576e

dflash: refactor draft model conversion (#25110)

ruixiang63 committed 5 days ago

Verified fa72bc68

chat : implement minicpm5 parser (#24889)

aldehir committed 5 days ago

Verified c818263f

jinja: add --dump-prog for debugging (#25086)

ngxson committed 5 days ago

Verified f68a788b

spec : add DFlash support (#22105)

ruixiang63 committed 5 days ago

Verified d1b34251

common : allow --offline in llama download (#25091)

angt committed 5 days ago

Verified c1a1c8ee

logs : reduce v2 (#25078)

ggerganov committed 5 days ago

Verified 27c8bb4f

Older