ggml-org/llama.cpp

Pull Requests Commits

remove event pending stage

0cc4m committed 105 days ago

0776a6a0

0cc4m committed 106 days ago

937a4256

fix event reuse issue with multiple vectors

0cc4m committed 106 days ago

5c177a10

NO MERGE: sync logging

0cc4m committed 107 days ago

ccd8d4a6

use multiple events to avoid reset issues

0cc4m committed 107 days ago

4374b5ab

don't use initializer list for semaphore wait info

0cc4m committed 107 days ago

eebf21c3

use timeline semaphores instead of fences for event_synchronize

0cc4m committed 107 days ago

08a4ba6f

also reset command buffers before reuse

0cc4m committed 107 days ago

2204bced

fix event command buffer reset validation error

0cc4m committed 107 days ago

c0d100e0

vulkan: fix event wait submission, event command buffer reset

0cc4m committed 107 days ago

58deae17

metal : fix l2 norm scale (#20493)

ggerganov committed 107 days ago

Verified 73c9eb8c

convert : fix/suppress pyright errors (#20442)

danbev committed 108 days ago

Verified 983df142

llama : disable graph reuse with pipeline parallelism (#20463)

ggerganov committed 108 days ago

Verified 57819b8d

vendor : update cpp-httplib to 0.37.1 (#20390)

cabelo committed 108 days ago

Verified 557fe2d9

tests : use `reasoning` instead of `reasoning_budget` in server tests (#20432)

pwilkin committed 108 days ago

Verified 0e810413

test-backend-ops: allow loading tests from file and parsing model operators into file (#19896)

0cc4m committed 108 days ago

Verified 128142fe

common : update completion executables list [no ci] (#19934)

danbev committed 108 days ago

Verified 6de1bc63

grammar: Fix grammar root symbol check (#19761)

AsbjornOlling committed 108 days ago

Verified 0a10c34d

vulkan: add GATED_DELTA_NET op support (#20334)

ProgenyAlpha committed 108 days ago

Verified deee2386

convert : better mtp check and fix return [no ci] (#20419)

CISC committed 108 days ago

Verified c3e3f9e5

vulkan: fix SSM_CONV PP scaling with large ubatch sizes (#20379)

ProgenyAlpha committed 108 days ago

Verified 40c550d4

New conversations now auto-select the first loaded model (#20403)

ServeurpersoCom committed 108 days ago

Verified de190154

ggml-virtgpu: Fix some build commands (#20341)

yomaytk committed 108 days ago

Verified 05039967

metal : avoid divisions in bin kernel (#20426)

ggerganov committed 108 days ago

Verified e4cff095

ci: Setup self-hosted CI for Intel Linux Vulkan backend (#20154)

rillomas committed 109 days ago

Verified 4cc6eb15

vulkan: fix l2_norm epsilon handling (#20350)

jeffbolznv committed 109 days ago

Verified 246ffc4b

vulkan: fix OOB check in flash_attn_mask_opt (#20296)

jeffbolznv committed 109 days ago

Verified aa429cf5

vulkan: Fix ErrorOutOfHostMemory on Intel GPU when loading large models with --no-mmap (#20059)

rillomas committed 109 days ago

Verified 5866e3bb

opencl: use larger workgroup size for get_rows (#20316)

lhez committed 109 days ago

Verified 0516e04b

opencl: add cumsum op (#18981)

shaofeiqi committed 109 days ago

Verified 3d9ab225

Older