ngxson/llama.cpp

Pull Requests Commits

convert mistral small

ngxson committed 278 days ago

00f0f8af

llama-tts : add '-o' option (#12398)

marcoStocchi committed 281 days ago

Verified f4c3dd5d

SYCL: Delete redundant plus sign and space (#12391)

aubreyli committed 281 days ago

Verified 3d35d87b

SYCL : support non-contiguous tensors in binary ops (add, sub, etc) (#12399)

fairydreaming committed 281 days ago

Verified b19bd064

[CANN]MUL_MAT optimization (#12382)

noemotiovon committed 281 days ago

Verified 92a39132

Add CLI arg to llama-run to adjust the number of threads used (#12370)

ericcurtin committed 282 days ago

Verified 9f2250ba

main : add -sysf / --system-prompt-file (#12249) (#12250)

CISC committed 282 days ago

Verified 774973b8

Load all MoE experts during warmup (#11571)

fairydreaming committed 282 days ago

Verified 8fcb5636

server: fix "--grammar-file" parameter (#12285)

dodekapod committed 282 days ago

Verified add2a3aa

graph : simplify attn input build for unified KV cache (#12381)

ggerganov committed 282 days ago

Verified c522ce41

hparams : add SWA rope parameters (#12374)

ggerganov committed 282 days ago

Verified 081bee8c

llama : fix Gemma3 SWA KV cache shift (#12373)

ggerganov committed 283 days ago

Verified 84d54755

arg : no n_predict = -2 for examples except for main and infill (#12364)

ngxson committed 283 days ago

Verified be7c3034

llama : refactor llama_context, llama_kv_cache, llm_build_context (#12181)

ggerganov committed 283 days ago

Verified e0dbec0b

server : fix crash when using verbose output with input tokens that are not in printable range (#12178) (#12338)

ishaangandhi committed 283 days ago

Verified 2048b591

Update build.yml for Windows Vulkan builder to use Vulkan 1.4.304 SDK for VK_NV_cooperative_matrix2 support (#12301)

oscarbg committed 283 days ago

Verified f08f4b31

llama.swiftui : fix xcframework dir in README [no ci] (#12353)

danbev committed 284 days ago

Verified 80a02aa8

sycl : variable sg_size support for mmvq kernels (#12336)

Alberto Cabrera Pérez committed 284 days ago

Verified 363f8c5d

CUDA/HIP: Fix fattn-vec-* when device warp size is not 32 (#12315)

IMbackK committed 284 days ago

Verified 34c961b1

llama : Add Gemma 3 support (+ experimental vision capability) (#12343)

ngxson committed 284 days ago

Verified 7841fc72

vulkan: fix bug in coopmat1 mul_mat_id (#12316)

jeffbolznv committed 284 days ago

Verified bf69cfe6

CUDA/HIP: refractor mmqv to unify the calculation of nwarps and rows per block between host and device code. (#12177)

IMbackK committed 284 days ago

Verified 10f2e818

ggml-backend : fix backend search path (#12330)

jklincn committed 285 days ago

Verified ba765438

metal : Cache the Metal library at the device context level (#12265)

BB-fat committed 285 days ago

Verified 6ab2e476

clip : bring back GPU support (#12322)

ngxson committed 285 days ago

Verified 96e12808

mat vec double buffer (#12188)

netrunnereve committed 285 days ago

Verified 2c9f833d

musa: support new arch mp_31 and update doc (#12296)

yeahdongcn committed 286 days ago

Verified 25136454

opencl: use OpenCL C standard supported by the device (#12221)

Henry Linjamäki committed 286 days ago

Verified 8acdacb3

readme: added Sidekick to available UIs (#12311)

johnbean393 committed 286 days ago

Verified 89b2b56e

tests : fix test-quantize-fns to init the CPU backend (#12306)

ggerganov committed 286 days ago

Verified e128a1bf

Older