ngxson/llama.cpp

Pull Requests Commits

ci : add caching for ROCm installation in release workflow (#15924)

danbev committed 144 days ago

Verified 33daece8

tests : filter out no-ops from coverage report (#15900)

danbev committed 144 days ago

Verified e7b6d83b

media : add transparent icon svg and png [no ci] (#15891)

06kellyjac committed 144 days ago

Verified 2cfef4d1

gitignore : Ignore vim swap files in tests (#15901)

createthis committed 144 days ago

Verified 09e72a03

CANN: Add ROPE sin/cos cache for reuse (#15912)

noemotiovon committed 144 days ago

Verified 10d8b2b6

CANN: implement LRU cache for ACL graphs (#15814)

noemotiovon committed 144 days ago

Verified 28b5f190

llama : check returned fn ptrs from ggml_backend_reg_get_proc_address (#15893)

danbev committed 144 days ago

Verified 86587da0

ci : cache ROCm installation in windows-latest-cmake-hip (#15887)

danbev committed 144 days ago

Verified ff02caf9

vulkan: throw the oom error instead of no memory type found (#15905)

0cc4m committed 145 days ago

Verified ae355f6f

vulkan: Fix OOB accesses in soft_max_back (#15861)

jeffbolznv committed 145 days ago

Verified 4f63cd70

HIP: use v_dot2_f32_f16 instruction for FA (#15884)

JohannesGaessler committed 145 days ago

Verified 17bc5a81

Workaround for subgroup arithmetic failing on MoltenVK with AMD GPUs (issue 15846) (#15886)

lksj92hs committed 145 days ago

Verified ed54e325

CUDA: Add mul_mat_id support for the mmf kernel (#15767)

am17an committed 145 days ago

Verified a972faeb

CUDA: fix GET_ROWS for large tensors (#15882)

JohannesGaessler committed 145 days ago

Verified 550cf726

contrib : add notes about merging PRs (#15881)

ggerganov committed 145 days ago

Verified c252ce67

requirements : update transformers/torch for Embedding Gemma (#15828)

danbev committed 145 days ago

Verified 70cd37db

model-conversion : add extra debugging support for model conversion (#15877)

pwilkin committed 145 days ago

Verified acc1b008

json : support `enum` values within `allOf` (#15830)

aldehir committed 146 days ago

Verified 7057faf6

media : add llama1 icon (#15878)

06kellyjac committed 146 days ago

Verified fe1c92cd

vulkan: sort graph to allow more parallel execution (#15850)

jeffbolznv committed 146 days ago

Verified e68aa10d

CUDA: generate_cu_files.py - add missing mxfp4 (#15880)

am17an committed 146 days ago

Verified 0a16bf52

chat : Deepseek V3.1 reasoning and tool calling support (OpenAI Style) (#15533)

createthis committed 146 days ago

Verified 88021565

server : bring back timings_per_token (#15879)

ngxson committed 146 days ago

Verified 56920f56

cuda : fix supports_op condition for get_rows when number of blocks is too large (#15868)

ggerganov committed 146 days ago

Verified b0d52998

metal : refactor + optimize (#15857)

ggerganov committed 146 days ago

Verified f28d4f4a

ggml: allow casting between f32 and i32 (#15783)

ngxson committed 146 days ago

Verified 9fcb29f2

CUDA: non-contiguous src0 not supported for PAD (#15869)

CISC committed 146 days ago

Verified 5ef22d28

convert : force setting sliding_window from original config (#15867)

danbev committed 146 days ago

Verified 233d773d

batched-bench : fix llama_synchronize usage during prompt processing (#15835)

ggerganov committed 146 days ago

Verified a885dcff

context : fix n_outputs during reserve (#15858)

ggerganov committed 146 days ago

Verified 663027fd

Newer Older