ngxson/llama.cpp

Pull Requests Commits

bring back _get_request_headers

ngxson committed 1 year ago

e8b7d263

Revert "multithreaded download"

ngxson committed 1 year ago

2e535f6f

ngxson committed 1 year ago

c8760ccb

multithread: print debug

ngxson committed 1 year ago

63f0604a

multithreaded download

ngxson committed 1 year ago

42fc895a

convert : fix flake8 lint regarding lamdba assigment

compilade committed 1 year ago

4c0170e2

convert : use writeable buffer for remote lazy tensors

compilade committed 1 year ago

78094fc4

support HF_TOKEN

ngxson committed 1 year ago

b584e397

small style fix

ngxson committed 1 year ago

4f657627

correct metadata.name

ngxson committed 1 year ago

df95a3a7

convert : allow using lazy remote tensors

compilade committed 1 year ago

3a3682de

convert: add --remote option

ngxson committed 1 year ago

08ecbbe3

ngxson committed 1 year ago

7f61d0b2

gguf util : add SafetensorRemote

ngxson committed 1 year ago

2507c8e2

Revert "sycl:remove redundant memcopy in function ggml_backend_sycl_buffer_set_tensor" (#12812)

NeoZhangJianyu committed 1 year ago

Verified 656babd6

gguf-py : support lazy tensor splitting (#12809)

compilade committed 1 year ago

Verified a226bc7a

llama : Support llama 4 text-only (#12791)

ngxson committed 1 year ago

Verified 1466621e

opencl: better identify Adreno GPU (#12760)

lhez committed 1 year ago

Verified 82974011

hellaswag: display estimated score confidence interval (#12797)

stduhpf committed 1 year ago

Verified 4ccea213

cuda : fix HIP and MUSA BF16 (#0)

ggerganov committed 1 year ago

1a1ab7e7

ggerganov committed 1 year ago

a4e46e28

ggml : simplify Arm fp16 CPU logic (ggml/1177)

ggerganov committed 1 year ago

ff067dbc

CUDA: don't convert BF16 weights to FP32 (ggml/1174)

CISC committed 1 year ago

36ca8b36

cpu: move all the operators into a separate c++ file (except mul_mat) (ggml/1167)

cmdr2 committed 1 year ago

995083e4

sycl: remove redundant memcopy in function ggml_backend_sycl_buffer_set_tensor (#12734)

zhouwg committed 1 year ago

Verified 518a0148

ci : no curl on ggml-ci (#12796)

ngxson committed 1 year ago

Verified e391d3ee

cmake : enable curl by default (#12761)

ngxson committed 1 year ago

Verified bd3f59f8

CANN: fix typo in ggml-cann (#12733)

zhouwg committed 1 year ago

Verified 52b3d71f

CANN: Refactor to reduce duplicate code (#12731)

hipudding committed 1 year ago

Verified d0d5b223

musa: fix compilation warnings in mp_22/31 (#12780)

yeahdongcn committed 1 year ago

Verified 916c83bf

Older