ngxson/llama.cpp

Pull Requests Commits

add clarify comment for norm eps

ngxson committed 1 year ago

Verified 1c0a25ba

restore boi/eoi check

ngxson committed 1 year ago

321d4ed1

add to list of pre-quant

ngxson committed 1 year ago

d0adae42

add test for internvl

ngxson committed 1 year ago

95be3888

ngxson committed 1 year ago

abb12389

rm mobilevlm from test

ngxson committed 1 year ago

6472d8c8

ngxson committed 1 year ago

ae25cd29

InternVL3-1B working

ngxson committed 1 year ago

2bfa7a4c

convert : internvl support

ngxson committed 1 year ago

e309f161

CUDA: fix FlashAttention on Turing (#13415)

JohannesGaessler committed 1 year ago

Verified d8919424

arg : add env var to control mmproj (#13416)

ngxson committed 1 year ago

Verified 7fef1176

vulkan: scalar flash attention implementation (#13324)

jeffbolznv committed 1 year ago

Verified dc1d2adf

chore(llguidance): use tagged version that does not break the build (#13413)

HRKings committed 1 year ago

Verified 7c28a74e

server : vision support via libmtmd (#12898)

ngxson committed 1 year ago

Verified 33eff402

sycl : implementation of reordered Q4_0 MMVQ for Intel GPUs (#12858)

Alberto Cabrera Pérez committed 1 year ago

Verified 17512a94

metal : optimize MoE for large batches (#13388)

ggerganov committed 1 year ago

Verified 611aa914

CUDA: FA support for Deepseek (Ampere or newer) (#13306)

JohannesGaessler committed 1 year ago

Verified 0cf6725e

llama : do not crash if there is no CPU backend (#13395)

slaren committed 1 year ago

Verified 27ebfcac

CUDA: fix crash on large batch size for MoE models (#13384)

JohannesGaessler committed 1 year ago

Verified 5c86c9ed

imatrix : Add --parse-special for enabling parsing of special tokens in imatrix calculation (#13389)

bartowski1182 committed 1 year ago

Verified efb8b47e

llama-run: add support for downloading models from ModelScope (#13370)

yeahdongcn committed 1 year ago

Verified 0527771d

mtmd : fix batch_view for m-rope (#13397)

ngxson committed 1 year ago

Verified 2189fd3b

llama : one-off chat template fix for Mistral-Small-2503 (#13398)

ngxson committed 1 year ago

Verified 3f96aeff

rpc : add rpc_msg_set_tensor_hash_req (#13353)

rgerganov committed 1 year ago

Verified b486ba05

vulkan: Allow up to 4096 elements for mul_mat_id row_ids (#13326)

jeffbolznv committed 1 year ago

Verified 02115dcd

server : (webui) rename has_multimodal --> modalities (#13393)

ngxson committed 1 year ago

Verified d9c4acca

ci : limit write permission to only the release step + fixes (#13392)

slaren committed 1 year ago

Verified 15e03282

mtmd : Expose helper_decode_image_chunk (#13366)

mattjcly committed 1 year ago

Verified f05a6d71

server : (webui) fix a very small misalignment (#13387)

ngxson committed 1 year ago

Verified ee01d71e

server : (webui) revamp the input area, plus many small UI improvements (#13365)

ngxson committed 1 year ago

Verified 8c83449c

Older