ngxson/llama.cpp

Pull Requests Commits

support vision input for mtp

ngxson committed 3 days ago

a748c270

Merge branch 'xsn/llama_batch_ext' into xsn/llama_batch_ext_2

ngxson committed 3 days ago

ca70c5d0

ngxson committed 3 days ago

c5ae143e

rm some stale calls

ngxson committed 3 days ago

52c0f8d4

server_batch to use common_batch

ngxson committed 3 days ago

da605de1

common_speculative_process

ngxson committed 3 days ago

798d2bff

ngxson committed 3 days ago

68964c95

Merge branch 'xsn/llama_batch_ext' into xsn/llama_batch_ext_2

ngxson committed 3 days ago

074d2b2f

upstream some changes

ngxson committed 3 days ago

e67da541

Merge remote-tracking branch 'upstream/master' into xsn/llama_batch_ext

ngxson committed 3 days ago

4dd6864c

ngxson committed 5 days ago

3477e17c

scripts : use sed instead of grep for version parsing [no ci] (#28700)

danbev committed 5 days ago

Verified 41fc7584

models: clean up some dead switch branches in old models (#28669)

iggy committed 5 days ago

Verified d344123f

tests : increase tolerance for Add fusion tests (#28691)

gaugarg-nv committed 5 days ago

Verified c32d1dab

tests : drop SYCL special-casing in test-backend-ops.cpp (#28688)

ggerganov committed 5 days ago

Verified e5a8d439

vulkan: fall back to shared-memory reduction for dmmv on PowerVR (#28341)

JulianPscheid committed 5 days ago

Verified 3ff67eb4

convert : expand Nemotron H conversion fix (#28689)

frozenblade1224 committed 5 days ago

Verified 8c322d5b

memory : avoid allocating V cache for indexer (it's not used) (#28330)

fairydreaming committed 5 days ago

Verified 311d4211

vulkan : add command-buffer debug labels for GPU profilers (#28101)

CordellBlanchard committed 5 days ago

Verified 72797e89

ggml-cpu(s390x): add repack support for q4_0 (#28667)

taronaeo committed 5 days ago

Verified 4ea6d1bb

ggml-cpu(s390x): add Q1_0 vector intrinsic support (#28606)

taronaeo committed 5 days ago

Verified f1b6fbf3

model: fix all granite family parameter counts (#28643)

taronaeo committed 5 days ago

Verified d7e86430

ci: fix sanitizer tests (#28583)

netrunnereve committed 6 days ago

Verified 434ddbbc

vulkan: use spec constant for matrix matrix multiplication A-type (#25773)

0cc4m committed 6 days ago

Verified 91f6a6cf

hexagon: rope updates (#28628)

tboinovski1 committed 6 days ago

Verified 6d9c82ea

vulkan: Convert FILL to distribute workgroups in 2D to avoid exceeding maxComputeWorkGroupCount (#28592)

rillomas committed 6 days ago

Verified 22397c31

py : lower numpy to 2.2.6 (#28654)

CISC committed 6 days ago

Verified 4b98ab80

py : bump numpy to 2.4.6 (#28649)

CISC committed 6 days ago

Verified 9cf3bf25

llama : use int32_t for llama_sampler_chain_n return type (#28631)

linsen458-spec committed 6 days ago

Verified 4850c772

CUDA: size routed MoE MMQ N-tiles from typical expert width on RDNA3 (#28552)

pwilkin committed 6 days ago

Verified d4abd573

Older