ggerganov/llama.cpp

Pull Requests Commits

llama : more metal-friendly KV cache PAD

ggerganov committed 1 year ago

Verified 33a004e9

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

llama-bench : add pp+tg test type (#7199)

slaren committed 1 year ago

Verified e8496488

metal : fix flash attention kernel requirements (#7169)

ggerganov committed 1 year ago

Verified 18e43766

convert : print "ignore_merges" field

ggerganov committed 1 year ago

Verified 8c660242

llama : use n_vocab to differentiate between mistral 7B and llama3 8B (#7200)

slaren committed 1 year ago

Verified 25c6e82e

Fix memory bug in grammar parser (#7194)

jart committed 1 year ago

Verified 4e388097

Main+: optionally allow special tokens from user in interactive mode (#7097)

hanishkvc committed 1 year ago

Verified f89fe273

Older