ggerganov/llama.cpp

Pull Requests Commits

llama: properly handle KV on training

ngxson committed 6 days ago

3f620574

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

JoursBleu committed 7 days ago

Verified 465e49b9

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

danielhanchen committed 7 days ago

Verified 5fdfa628

[Model] Support for Spark2_5ForCausalLM implementation (#27868)

KnightYao committed 7 days ago

Verified 3ad1ba73

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

lhez committed 7 days ago

Verified d03efa5d

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

am17an committed 7 days ago

Verified 73a43d1f

grammar : fix max repetition threshold (#28469)

aldehir committed 7 days ago

Verified 9e0e2205

ui: Improve Chat Messages rendering performance (#28460)

allozaur committed 7 days ago

Verified 0afb805b

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

ngxson committed 7 days ago

Verified 7620399f

ui : embed assets directly with CMake (#28445)

angt committed 7 days ago

Verified c457e3bf

metal : add remaining fa-vec tunings for M2 Max (#28458)

nikwen committed 7 days ago

Verified 971595d6

Github: limit blank issues to maintainers (#28435)

JohannesGaessler committed 7 days ago

Verified 74a7c897

metal : fix memory leak in early return (#28399)

nikwen committed 8 days ago

Verified 6a1a922d

sycl : fix test-backend-ops CI break && restore Kronecker product FWHT support (#28016) (#28254)

philip-jingxin committed 8 days ago

Verified 4d917609

sycl: attribute device allocations by site (GGML_SYCL_MEMTRACE) (#27631)

nicois committed 8 days ago

Verified cd8cdf39

metal : add remaining fa-vec tunings for M3 (#28396)

init-22 committed 8 days ago

Verified 427291b5

ui : fix MCP image attachments not displayed in tool block (#25789) (#28089)

nachobh committed 8 days ago

Verified 85d5703a

opencl: extend the elementwise and data‐movement op coverage (#27633)

wanghqc committed 9 days ago

Verified 1548a240

opencl: add Adreno xmem SDPA path (#26331)

happyyzy committed 9 days ago

Verified 4acf4a4c

ci : move more jobs to ccache-buckets (#28375)

CISC committed 9 days ago

Verified 8b4b3558

ui: export conversations from database instead of cached store (#27432)

intel00000 committed 9 days ago

Verified 1863ac03

model : add Tencent Hy 4 (hy_v4) preview architecture support (#28127)

Little0o0 committed 9 days ago

Verified 49c0dc82

llama.cpp : bump version to 0.4.0 (#28386)

ggerganov committed 9 days ago

Verified 5266f24d

sync : ggml (#28379)

ggerganov committed 9 days ago

Verified 64a155d2

model, mtmd: fix gemma4 vision handling (#28335)

ngxson committed 9 days ago

Verified 163a4079

metal : add remaining fa-vec tunings for M3 Max (#28373)

nikwen committed 9 days ago

Verified 8f83678f

ggml : replace compile definitions with version.h.in (#28364)

danbev committed 9 days ago

Verified 86b351fd

Don't use npx inside a package.json script (#28270)

eapache committed 9 days ago

Verified d509cb1e

ggml : don't crash when backend search path can't be read (#28271)

angt committed 9 days ago

Verified 4cbe8b07

ggml : remove GGML_CUDA_PEER_MAX_BATCH_SIZE (#28177)

angt committed 9 days ago

Verified 24f5bf8a

Older