ggml-org/llama.cpp

Pull Requests Commits

cont : report malloc failure

ggerganov committed 17 days ago

Verified d211167a

ggml : add get_alloc_size_n to buffer type interface

ggerganov committed 17 days ago

Verified fbfe6472

ggml : address review comments on alloc_buffer_n

ggerganov committed 17 days ago

Verified eb6d54a0

tests : add descriptions for alloc_buffer_n tests

ggerganov committed 17 days ago

Verified 5391208d

cont : fix compile warnings

ggerganov committed 17 days ago

Verified de9cb4f8

tests : add alloc_buffer_n coverage

ggerganov committed 17 days ago

Verified 08679c60

ggml : add TODO tag for shared buffer split logic

ggerganov committed 17 days ago

Verified 5980ad6a

cont : fix `cur_buf_size` init after flushing a buffer

ggerganov committed 17 days ago

Verified 4ba972aa

ggml : add `alloc_buffer_n` to buffer type interface

ggerganov committed 17 days ago

Verified 59c277e3

pyproject : add linux platform marker to uv torch source (#29177)

Yezat committed 18 days ago

Verified 207bdab9

hexagon: install rebuilt HTP skels (#29828)

kurquhar committed 18 days ago

Verified 5fc4f3c8

qwen4exp: fix tests (#29819)

am17an committed 18 days ago

Verified 159c651f

hexagon: add q2_k and q3_k quant type support (#29717)

jhen0409 committed 18 days ago

Verified a868c3e3

CUDA: fix 2 broken Volta FA cases (#29803)

JohannesGaessler committed 18 days ago

Verified ec7630a6

llama: refer to segment documentation [no ci] (#29074)

JohannesGaessler committed 18 days ago

Verified 78e2964c

common,rpc : fix cache dir creation through symlinks on buggy libstdc++ (#29816)

angt committed 18 days ago

Verified f1cee994

skill: note about model-specific CLI arguments + testings (#29808)

ngxson committed 18 days ago

Verified 68e79bd8

ci: fix Fusion / metal by updating the qwen4exp baseline (#29812)

ServeurpersoCom committed 18 days ago

Verified e358d591

llama : clamp kpool re-pool bound to existing pools (#29805)

ggerganov committed 18 days ago

Verified 81e39ad3

hexagon: shared strided DMA copy for CPY and CONCAT, any-dim CONCAT via DMA (#29685)

njsyw1997 committed 18 days ago

Verified dcd387a4

server: return HTTP 400 for invalid embedding requests (#29060)

SamMalayek committed 18 days ago

Verified d775ebf3

convert : write Gemma embedding scale for DFlash drafts (#29802)

kabu1204 committed 18 days ago

Verified 2b36825c

cuda : route sm70 to the Turing MMVQ nwarps table (#29753)

tkittich committed 18 days ago

Verified 42d95816

metal : release temporary private transfer buffers (#29777)

mvanlamz committed 18 days ago

Verified 13b4d713

webgpu: add bfloat16 support for MUL_MAT/MUL_MAT_ID/GET_ROWS- #29358 (#29358)

yomaytk committed 18 days ago

Verified 4b1622af

llama : fix invalid assert in recurrent memory (#29799)

ggerganov committed 18 days ago

Verified 869034b4

CUDA: Handle compute type for NVFP4 on cublass path (#29173)

ynankani committed 18 days ago

Verified b56f34ab

Qwen4Exp: add MTP (#29761)

am17an committed 18 days ago

Verified c061df19

llama: fix qwen4exp (#29751)

am17an committed 18 days ago

Verified 66e0c17e

CUDA: Make CCCL configurable + pin it to 3.4.3 for CI jobs (#29792)

ORippler committed 18 days ago

Verified 76776785

Older