ggml-org/llama.cpp

Pull Requests Commits

server : avoid context swaps by shifting the KV cache

ggerganov committed 2 years ago

Verified c5650ed4

server : clear the KV cache beyond n_past before llama_decode

ggerganov committed 2 years ago

Verified ce2d995a

examples : do not eval prompt 2 times (close #3348)

ggerganov committed 2 years ago

Verified 2b8830af

examples : add example for batched decoding

ggerganov committed 2 years ago

Verified a2075615

examples : utilize new llama_get_logits_ith()

ggerganov committed 2 years ago

Verified d008733e

metal : use mm kernels for batch size > 2

ggerganov committed 2 years ago

Verified 4c72ab13

llama : simplify returns if/else branches

ggerganov committed 2 years ago

Verified e9463792

parallel : fix crash when `-n -1`

ggerganov committed 2 years ago

Verified 4ad06769

Merge branch 'master' into custom-attention-mask

ggerganov committed 2 years ago

Verified 25856900

readme : add Mistral AI release 0.1 (#3362)

paschembri committed 2 years ago

Verified 4aea3b84

ggml-cuda : perform cublas fp16 matrix multiplication as fp16 (#3370)

slaren committed 2 years ago

Verified da040034

convert : remove bug in convert.py permute function (#3364)

jzhang38 committed 2 years ago

Verified e5196210

make-ggml.py : compatibility with more models and GGUF (#3290)

richardr1126 committed 2 years ago

Verified ac435761

gguf : fix a few general keys (#3341)

cebtenzzre committed 2 years ago

Verified 20c7e1e8

metal : reusing llama.cpp logging (#3152)

Ricardicus committed 2 years ago

Verified dc689740

build : add ACCELERATE_NEW_LAPACK to fix warning on macOS Sonoma (#3342)

jagtesh committed 2 years ago

Verified 527e57cf

readme : add some recent perplexity and bpw measurements to READMES, link for k-quants (#3340)

BarfingLemurs committed 2 years ago

Verified ffe88a36

llama : fix kv cache heuristic when context is less than 32

ggerganov committed 2 years ago

Verified c1596f63

cmake : fix build-info.h on MSVC (#3309)

dranger003 committed 2 years ago

Verified 99115f3f

docs: Fix typo CLBlast_DIR var. (#3330)

fossilet committed 2 years ago

Verified 1726f962

nix : add cuda, use a symlinked toolkit for cmake (#3202)

Green-Sky committed 2 years ago

Verified a98b1633

llama-bench : add README (#3317)

slaren committed 2 years ago

Verified c091cdfb

examples : fix RoPE defaults to match PR #3240 (#3315)

cebtenzzre committed 2 years ago

Verified 51a7cf5c

scripts : use `/usr/bin/env` in shebang (#3313)

kevinji committed 2 years ago

Verified bedb92b6

Update README.md (#3289)

leedrake5 committed 2 years ago

Verified bc9d3e39

ggml-opencl.cpp: Make private functions static (#3300)

shibe2 committed 2 years ago

Verified 36b904e2

simple : add README.md

ggerganov committed 2 years ago

Verified 88451600

llama : llama.h formatting + comments

ggerganov committed 2 years ago

Verified 5a3369d8

zig : fix for updated c lib (#3259)

edeetee committed 2 years ago

Verified 324f3403

embedding : update README.md (#3224)

yuiseki committed 2 years ago

Verified f56c418a

Older