ggerganov/llama.cpp

Pull Requests Commits

llama-vocab : add SuperBPE pre-tokenizer

compilade committed 200 days ago

35c2f8b9

server : Add verbose output to OAI compatible chat endpoint. (#12246)

mglambda committed 200 days ago

Verified 77f9c6bb

install : add macports (#12518)

IOOI-SqAR committed 200 days ago

Verified 18b663d8

llama : gemma3 : use output tensor if it exists in model weight (#12506)

ngxson committed 201 days ago

Verified fbdfefe7

ggml : fix quantized cpy op (#12310)

ggerganov committed 201 days ago

Verified ba932dfb

musa: refine compute capability (#12493)

yeahdongcn committed 201 days ago

Verified fac63a3d

vulkan: Optimize mul_mat_vec p021 and nc shaders (#12505)

jeffbolznv committed 201 days ago

Verified eddfb438

Vulkan: RTE rounding for cpy to quant (#12480)

stduhpf committed 202 days ago

Verified 4375415b

vulkan: workaround for AMD Windows driver 16 bit unpack8 bug (#12472)

netrunnereve committed 202 days ago

Verified 30c42ef5

model : do not repack if a GPU device is present (#12498)

ggerganov committed 202 days ago

Verified af04481e

chore : cleanup llama_model_loader::TENSOR_ usage (#12492)

CISC committed 202 days ago

Verified 960e7260

llama-tts : avoid crashes related to bad model file paths (#12482)

marcoStocchi committed 202 days ago

Verified ea1518e8

[SYCL] Fix build on Windows when ccache enabled (#9954) (#9976)

MakeDecisionWorth committed 202 days ago

Verified 1aa87ee5

sycl: cleanup oneDNN related code (#12097)

sgeor255 committed 203 days ago

Verified 9ffcc9e3

webui : Prevent rerendering on textarea input (#12299)

woof-dog committed 203 days ago

Verified e0464306

llama : make Qwen2MoE QKV bias optional (#12477)

CISC committed 203 days ago

Verified dbb3a473

ggml : block interleaving support for Q4_K quantization for x86 AVX2 architecture (#12332)

Srihari-mcw committed 203 days ago

Verified 3d82dbcb

convert : avoid calls to tokenizer.added_tokens_decoder (#12473)

bartowski1182 committed 203 days ago

Verified 732b5fbf

context : clear sets containing encoder output sequence ids before storing new values (#12470)

fairydreaming committed 204 days ago

Verified 568013d0

CUDA: Improve flash decoding kernel GPU occupancy for BS=1 case (#12183)

gaugarg-nv committed 204 days ago

Verified 517b5ddb

vulkan: optimize iq1 coopmat2 dequant functions (#12427)

jeffbolznv committed 204 days ago

Verified a9b59288

Fix visionOS build and add CI (#12415)

guusw committed 204 days ago

Verified 0fd8487b

llama : add support for GPT2, Bloom and CodeShell tied word embeddings (#12456)

CISC committed 204 days ago

Verified 108e53c2

convert : Support chat_template.json (#12460)

CISC committed 204 days ago

Verified a686171e

vulkan: Submit once enough matmul work has been recorded (#12406)

jeffbolznv committed 204 days ago

Verified c446b2ed

opencl: improve profiling (#12442)

lhez committed 205 days ago

Verified d84635b1

graph : normalize Q, K, V shapes + sync cross attention (#12449)

ggerganov committed 205 days ago

Verified 75422e8b

musa: override warp_size of musa device to 32 (#12445)

yeahdongcn committed 205 days ago

Verified bb115d2b

llama : support converting Mistral Small text-only (#12450)

ngxson committed 205 days ago

Verified 29fff308

speculative : fix seg fault in certain cases (#12454)

ggerganov committed 205 days ago

Verified c6af2161

Older