ggml-org/llama.cpp

Pull Requests Commits

remove event pending stage

0cc4m committed 11 hours ago

a338a1e2

add semaphore wait condition also if compute_ctx already exists

0cc4m committed 1 day ago

d287bbbb

fix event reuse issue with multiple vectors

0cc4m committed 1 day ago

9adf514a

use multiple events to avoid reset issues

0cc4m committed 1 day ago

4374b5ab

don't use initializer list for semaphore wait info

0cc4m committed 2 days ago

eebf21c3

use timeline semaphores instead of fences for event_synchronize

0cc4m committed 2 days ago

08a4ba6f

also reset command buffers before reuse

0cc4m committed 2 days ago

2204bced

fix event command buffer reset validation error

0cc4m committed 2 days ago

c0d100e0

vulkan: fix event wait submission, event command buffer reset

0cc4m committed 2 days ago

58deae17

metal : fix l2 norm scale (#20493)

ggerganov committed 2 days ago

Verified 73c9eb8c

convert : fix/suppress pyright errors (#20442)

danbev committed 2 days ago

Verified 983df142

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

ggerganov committed 3 days ago

Verified 57819b8d

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

cabelo committed 3 days ago

Verified 557fe2d9

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

pwilkin committed 3 days ago

Verified 0e810413

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

0cc4m committed 3 days ago

Verified 128142fe

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

danbev committed 3 days ago

Verified 6de1bc63

grammar: Fix grammar root symbol check (#19761)

AsbjornOlling committed 3 days ago

Verified 0a10c34d

vulkan: add GATED_DELTA_NET op support (#20334)

ProgenyAlpha committed 3 days ago

Verified deee2386

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

CISC committed 3 days ago

Verified c3e3f9e5

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

ProgenyAlpha committed 3 days ago

Verified 40c550d4

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

ServeurpersoCom committed 3 days ago

Verified de190154

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

yomaytk committed 3 days ago

Verified 05039967

metal : avoid divisions in bin kernel (#20426)

ggerganov committed 3 days ago

Verified e4cff095

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

rillomas committed 3 days ago

Verified 4cc6eb15

vulkan: fix l2_norm epsilon handling (#20350)

jeffbolznv committed 3 days ago

Verified 246ffc4b

vulkan: fix OOB check in flash_attn_mask_opt (#20296)

jeffbolznv committed 3 days ago

Verified aa429cf5

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

rillomas committed 3 days ago

Verified 5866e3bb

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

lhez committed 3 days ago

Verified 0516e04b

opencl: add cumsum op (#18981)

shaofeiqi committed 3 days ago

Verified 3d9ab225

hip: compile debug builds with -O2 on hip to avoid a compiler bug (#20392)

IMbackK committed 3 days ago

Verified d63aa398

Older