ggerganov/llama.cpp

Pull Requests Commits

allow dmabuf p2p

0cc4m committed 22 days ago

af2146b7

fix compile issue

0cc4m committed 22 days ago

b4acc6d3

Merge branch 'pr-25051' into 0cc4m/vulkan-allreduce

0cc4m committed 22 days ago

6634f71f

reverse check order

0cc4m committed 22 days ago

e46b3ae5

fix compile issues

0cc4m committed 22 days ago

50e0a2dd

llama.cpp : bump version to 0.3.0 (#27696)

ggerganov committed 22 days ago

Verified c1d0e7a0

use memcpy for small copies across host-visible memory

0cc4m committed 22 days ago

4d879b9c

add semi-async staging copy and use it over syncfd for small copies

0cc4m committed 22 days ago

bce45fbe

swap dmabuf check order

0cc4m committed 22 days ago

bac13690

use allocator instead of hop2

0cc4m committed 22 days ago

86ad4417

use sync_fd binary semaphores for cross-driver synchronization

0cc4m committed 22 days ago

5f512ae7

double buffering

0cc4m committed 22 days ago

4f4cea31

fixes, disable semaphore sharing on Nvidia + non-Nvidia

0cc4m committed 22 days ago

3e53e38d

0cc4m committed 22 days ago

bf316ac5

vulkan: add optimized device to device copy function

0cc4m committed 22 days ago

740c1273

ggerganov committed 22 days ago

81191aff

ggml : bump version to 0.22.0 (ggml/1607)

ggerganov committed 22 days ago

93882361

grammar : parse \- in char classes as literal hyphen (#27591)

NIXKnight committed 23 days ago

Verified eb25b726

sycl : mark tq2_0 as not supported (#27660)

arthw committed 23 days ago

Verified 814d84bc

webgpu : fix handling of infinity values during ARGSORT and TOP_K (#27538)

fairydreaming committed 23 days ago

Verified 5ea87dda

metal : per-device tuned (Q, NE) for flash-attn vec (#26570)

forforever73 committed 23 days ago

Verified f280b269

metal: per-op source split + parallel compile (#26561)

forforever73 committed 23 days ago

Verified b615f5b4

misc : read repetition_penalty from generation_config.json (#27659)

tdakhran committed 23 days ago

Verified b3c3b96a

tests : disable DOTS3NOTE arch test for WebGPU (#27654)

fairydreaming committed 23 days ago

Verified 75844307

convert: fix GLM regression in index_tensors (#27655)

jacekpoplawski committed 23 days ago

Verified 71cc86fa

ggml : shorten virtual device naming in CUDA and Metal (#27608)

ggerganov committed 23 days ago

Verified a14dba68

webgpu : reorder includes since V that appears in common_decls.tmpl may be defined as K in flash_attn_decls.tmpl if KV_OVERLAP (#27545)

fairydreaming committed 23 days ago

Verified c1c766da

mtmd: video: fix moov atom at the end of file (#27596)

ngxson committed 23 days ago

Verified 160c6b0b

ci : apply ccache-clear with older/min/dry-run to all ccache jobs (#27602)

ggerganov committed 23 days ago

Verified 985b1491

ggml : fix ggml_clamp (#27644)

ggerganov committed 23 days ago

Verified 6036c635

Older