ggerganov/llama.cpp

Pull Requests Commits

convert : fix mistakes from refactoring

cebtenzzre committed 1 year ago

ccd757a1

cleanup convert-hf-to-gguf.py

cebtenzzre committed 1 year ago

c2f407e3

convert : pad vocab size to multiple of 64, not 8

cebtenzzre committed 1 year ago

b8ff85ef

cebtenzzre committed 1 year ago

48a7ef6e

llama : allow raw byte in SPM vocabs; don't crash on nl 404 (#5478)

akx committed 1 year ago

Verified c4e6dd59

llama : make load error reporting more granular (#5477)

akx committed 1 year ago

Verified 037259be

finetune : rename feed-forward tensors (w1/w2/w3) (#4839)

danbev committed 1 year ago

Verified 26397890

tests : multi-thread the tokenizer tests (#5474)

ggerganov committed 1 year ago

Verified cf45252a

llama : support batched embeddings (#5466)

iamlemec committed 1 year ago

Verified 03bf161e

make: add error message for bad CUDA version (#5444)

JohannesGaessler committed 1 year ago

Verified ad014bba

bert : add tests + fix quantization (#5475)

ggerganov committed 1 year ago

Verified 49cc1f7d

tests : disable moe test (#5473)

ggerganov committed 1 year ago

Verified 99b8b43d

ggml-quants : fix compiler warnings (shadow variable) (#5472)

ikawrakow committed 1 year ago

Verified 895407f3

llama : fix quantization when tensors are missing (#5423)

ggerganov committed 1 year ago

Verified 099afc62

swift : package no longer use ggml dependency (#5465)

ggerganov committed 1 year ago

Verified df334a11

py : fix persimmon `n_rot` conversion (#5460)

lx200916 committed 1 year ago

Verified dbd8828e

ggml-sycl: Replace 3d ops with macro (#5458)

abhilash1910 committed 1 year ago

Verified 43fe07c1

llava : remove prog parameter from ArgumentParser (#5457)

danbev committed 1 year ago

Verified 4a46d2b7

sync : ggml (#5452)

ggerganov committed 1 year ago

Verified 3b169441

CUDA: mul_mat_vec_q tiling, refactor mul mat logic (#5434)

JohannesGaessler committed 1 year ago

Verified 3bdc4cd0

Add support for BERT embedding models (#5423)

iamlemec committed 1 year ago

Verified 2891c8aa

flake.lock: Update

github-actions[bot] committed 1 year ago

97a33650

vulkan: only use M-sized matmul on Apple GPUs (#5412)

slp committed 1 year ago

Verified c88c74f9

common : use enums for sampler types (#5418)

z80maniac committed 1 year ago

Verified a803333a

server : allow to specify tokens as strings in logit_bias (#5003)

z80maniac committed 1 year ago

Verified 68478014

main : ctrl+C print timing in non-interactive mode (#3873)

ggerganov committed 1 year ago

Verified 85910c5b

common : fix compile warning

ggerganov committed 1 year ago

Verified 139b62a8

ggml : fix compile warnings (unused vars) (#4966)

ggerganov committed 1 year ago

Verified 0f2411f1

ggml : add mmla kernels for quantized GEMM (#4966)

snadampal committed 1 year ago

Verified a07d0fee

lookup: add print for drafting performance (#5450)

JohannesGaessler committed 1 year ago

Verified e4640d8f

Older