ggerganov/llama.cpp

Pull Requests Commits

ggerganov committed 31 days ago

Verified 85e8b2a4

vendor : update cpp-httplib to 0.57.1 (#29239)

angt committed 31 days ago

Verified ff0dbb97

metal : fix mask bounds in flash attention block pre-pass (#29220)

masterFoad committed 31 days ago

Verified fb34fc26

test-save-load-state : compare logits with NMSE and feed expected tokens (#29238)

ggerganov committed 31 days ago

Verified c641dfa8

llama-context : report graph inputs and input tensors during sched reserve (#26625)

ggerganov committed 31 days ago

Verified 96550613

cuda: fix sm_70 tile compilation error (#29224)

lingyezhixing committed 31 days ago

Verified b1c2863e

ggml-cuda : convert contiguous tensors four elements at a time (#29155)

pwilkin committed 31 days ago

Verified f4e276a2

cuda : accelerate conv2d with implicit GEMM (#29135)

leejet committed 31 days ago

Verified e6cef815

ggml : fix dimension and stride truncation in ggml_permute (#29227)

leejet committed 31 days ago

Verified c21284cd

vendor : update cpp-httplib to 0.57.0 (#29214)

angt committed 31 days ago

Verified 6f41ac59

docker : bump cuda to 13.4.1 (#29207)

CISC committed 31 days ago

Verified ec91ab5a

sycl : support gated DSV4_HC_PRE and optional HC_POST comb matrix (#29132)

cwriter committed 31 days ago

Verified bb3c853c

sycl : pinned memory use right device context instead of 0 (#28895)

lslusarczyk committed 31 days ago

Verified af911149

CUDA: Follow up of #25635, refactoring FA shared smem swizzle (#28536)

ynankani committed 31 days ago

Verified 1884824f

test-llama-archs : make tensor data stdev configurable and improve help (#29133)

ggerganov committed 31 days ago

Verified 161755f2

tests/test-backend-ops : allow regex entries in the -o filter (#29204)

ggerganov committed 31 days ago

Verified 1d72b05d

ci : refactor build-self-hosted into backend-specific workflows (#28991)

CISC committed 31 days ago

Verified 542e9202

args: add env vars for temperature, top-p, min-p and penalties (#27380)

kucharskim committed 31 days ago

Verified e0dff584

server : do not forward --api-key-file to router-spawned child instances (#28938)

nandan2003 committed 31 days ago

Verified 982a3329

tests : remove stale comment (#29140)

mostafafaheem committed 32 days ago

Verified 711f60be

ggml-metal : simplify fusion pattern op list declaration (#29206)

ggerganov committed 32 days ago

Verified 335b21fc

json: Fixed json enum handling (#28518)

Silverside committed 32 days ago

Verified 26394b4e

sycl : coalesce MKL-FA softmax loads instead of one work-item per row (#28918)

anantshri committed 32 days ago

Verified 1aa2954b

ggml-cpu: ARM Repack kernels for Q1_0 (#23492)

pl752 committed 32 days ago

Verified 8034c1d1

ci : Upgrade CUDA to 13.4 for Ubuntu CUDA Release Builds (#29202)

sam-india-007 committed 32 days ago

Verified 6ad1af56

hexagon: overhaul of buffer and DMA handling to support 64bit mappings + improvements (#29197)

max-krasnyansky committed 32 days ago

Verified 0c3626ec

cuda : tune MMVQ to MMQ crossover for SM70 (Volta) (#28912)

cyyself committed 32 days ago

Verified 68d9053a

metal : fix deprecation warnings from macOS 27 SDK (#29136)

nikwen committed 32 days ago

Verified 8aa161b5

webgpu : add fused gdn + cpy (#28976)

yomaytk committed 32 days ago

Verified 932a68e0

convert: enable --fuse-qkv for muse-glimmer (#29203)

dfriehs committed 32 days ago

Verified 62668d6b

Older