ggerganov/llama.cpp

Pull Requests Commits

examples : restore the functionality to import llama2.c models (#2685)

ochafik committed 2 years ago

Verified 95385241

fix convert-lora-to-ggml.py (#2738)

slaren committed 2 years ago

Verified 335acd2f

main : insert bos if no tokens (#2727)

klosax committed 2 years ago

Verified 5290c38e

gitignore : fix for windows (#2729)

akawrykow committed 2 years ago

Verified cc34dbda

chmod : make scripts executable (#2675)

cebtenzzre committed 2 years ago

Verified 7c2227a1

devops : RPM Specs (#2723)

jboero committed 2 years ago

Verified f19dca04

Fix values shown in the quantize tool help (#2735)

ikawrakow committed 2 years ago

Verified 8207214b

Strided perplexity (#2714)

ikawrakow committed 2 years ago

Verified 62959e74

Fix ggml to gguf conversion on Windows (#2733)

IgnacioFDM committed 2 years ago

Verified 7f7ddd50

server : allow json array in prompt or content for direct token input (#2306)

jxy committed 2 years ago

Verified b8ad1b66

docs : add grammar docs (#2701)

ejones committed 2 years ago

Verified f5fe98d1

Improve handling of special tokens in GGML to GGUF converter (#2725)

KerfuffleV2 committed 2 years ago

Verified 777f42ba

llama : fix whitespace escaping in tokenizer (#2724)

goerch committed 2 years ago

Verified 46ef5b5f

CUDA: use mul_mat_q kernels by default (#2683)

JohannesGaessler committed 2 years ago

Verified c63bb1d1

convert.py : clarifying error message (#2718)

apetenchea committed 2 years ago

Verified 3b6cfe7c

Fix CUDA softmax by subtracting max value before exp (#2665)

li-plus committed 2 years ago

Verified 800c9635

gguf : add ftype meta info to the model (#2710)

ggerganov committed 2 years ago

Verified deb7dfca

Quantization imrovements for k_quants (#2707)

ikawrakow committed 2 years ago

Verified bac66994

embedding : evaluate prompt in batches (#2713)

slaren committed 2 years ago

Verified 519c981f

ggml-cuda : use graph allocator (#2684)

slaren committed 2 years ago

Verified 1123f7fb

ggml : sync latest (SAM + SD operators, CUDA alibi) (#2709)

ggerganov committed 2 years ago

Verified ef3f333d

llama-bench : minor fixes (#2695)

slaren committed 2 years ago

Verified 8e4364f2

ggml : support CUDA's half type for aarch64(#1455) (#2670)

KyL0N committed 2 years ago

Verified 1e3bc523

metal : add missing barriers for mul-mat (#2699)

lshzh-ww committed 2 years ago

Verified 14b1d7e6

server : fallback to default if client param is null (#2688)

jhen0409 committed 2 years ago

Verified 226255b4

Fix convert-llama-ggmlv3-to-gguf.py vocab conversion (#2698)

KerfuffleV2 committed 2 years ago

Verified 930523c8

py : remove obsolete script

ggerganov committed 2 years ago

Verified c8dba409

gguf : new file format with flexible meta data (beta) (#2398)

ggerganov committed 2 years ago

Verified 6381d4e1

metal : fix synchronization in new matrix multiplication kernel (#2686)

lshzh-ww committed 2 years ago

Verified dadbed99

HellaSwag: split token evaluation into batches if needed (#2681)

ikawrakow committed 2 years ago

Verified cb1c0727

Newer Older