ggerganov/llama.cpp

Pull Requests Commits

Merge branch 'master' into fix-convert-modelname

mofosyne committed 1 year ago

Verified 284870c8

convert-hf : support direct Q8_0 conversion (#7234)

compilade committed 1 year ago

Verified ee522250

llama : less KV padding when FA is off (#7257)

ggerganov committed 1 year ago

Verified 614d3b91

llava-cli: fix base64 prompt (#7248)

Adriankhl committed 1 year ago

Verified 30e70334

perplexity: add BF16 vs. FP16 results (#7150)

JohannesGaessler committed 1 year ago

Verified 1c570d8b

[SYCL] rm wait() (#7233)

arthw committed 1 year ago

Verified 948f4ec7

llama : rename jina tokenizers to v2 (#7249)

Joan Fontanals committed 1 year ago

Verified 9aa67249

convert.py: Outfile default name change and additional metadata support (#4858)

mofosyne committed 1 year ago

Verified b1f8af18

change default temperature of OAI compat API from 0 to 1 (#7226)

Kartoffelsaft committed 1 year ago

Verified e586ee42

[SYCL] Add oneapi runtime dll files to win release package (#7241)

arthw committed 1 year ago

Verified cbf75894

[SYCL] update CI with oneapi 2024.1 (#7235)

arthw committed 1 year ago

Verified 0d5cef78

CUDA: add FP32 FlashAttention vector kernel (#7188)

JohannesGaessler committed 1 year ago

Verified dc685be4

cmake : fix version cmp (#7227)

ggerganov committed 1 year ago

Verified 6f1b6360

remove convert-lora-to-ggml.py (#7204)

slaren committed 1 year ago

Verified b228aba9

metal : fix warnings (skipme) (#0)

ggerganov committed 1 year ago

Verified 7bd4ffb7

ggerganov committed 1 year ago

Verified 1622ac02

metal : fix indent (ggml/0)

ggerganov committed 1 year ago

Verified 6aeff24f

ggml : resolve merge (ggml/0)

ggerganov committed 1 year ago

Verified 325756d2

Scripting & documenting debugging one test without anything else in the loop. (#7096)

josh-ramer committed 1 year ago

Verified fed01084

fix system prompt handling (#7153)

ngxson committed 1 year ago

Verified 72c177c1

convert-hf : support bfloat16 conversion (#7158)

compilade committed 1 year ago

Verified 5a419926

ggerganov committed 1 year ago

fae9d234

feat: implemented sigmoid function (ggml/806)

justcho5 committed 1 year ago

f5ef34e4

build: fix and ignore msvc warnings (ggml/805)

iboB committed 1 year ago

ef0d5e3e

convert : skip unaccessible HF repos (#7210)

CrispStrobe committed 1 year ago

Verified 3292733f

server : free llama_batch on exit (#7212)

stevegrubb committed 1 year ago

Verified 98863133

llama : lookup word in vocab before doing BPE merges (#7193)

tonyfettes committed 1 year ago

Verified f99e1e45

server: fix reported top tokens for temperature 0 (#7203)

JohannesGaessler committed 1 year ago

Verified 5ae3426b

llama : add Jina Embeddings architecture (#6826)

JoanFM committed 1 year ago

Verified b83cc3f5

ggml : full ALiBi support (#7192)

ggerganov committed 1 year ago

Verified 9cb317f7

Older