ggerganov/llama.cpp

Pull Requests Commits

metal : multi-simd softmax

ggerganov committed 1 year ago

Verified 46868a49

common : minor (#3715)

ggerganov committed 1 year ago

Verified ff8f9a88

llm : add llm_build_context (#3881)

ggerganov committed 1 year ago

Verified 50337961

common : allow caller to handle help/argument exceptions (#3715)

bandoti committed 1 year ago

Verified 0e40806c

log : make generating separate log files optional (#3787)

staviq committed 1 year ago

Verified a2758d08

sampling : null grammar field after reset (#3885)

l3utterfly committed 1 year ago

Verified e75dfdd3

ggml : fix UNUSED macro (#3762)

ggerganov committed 1 year ago

Verified 9a3b4f6c

finetune : add -ngl parameter (#3762)

AndrewGodfrey committed 1 year ago

Verified 73bdcb39

scripts : add server-llm.sh (#3868)

ggerganov committed 1 year ago

Verified f0e20932

server : re-enable completion and embedded at the same time (#3876)

a-h committed 1 year ago

Verified ca190bca

llama : refactor graph build code (#3837)

ggerganov committed 1 year ago

Verified 71e3718a

samplers : Min-P sampler implementation [alternative to Top P/Top K] (#3841)

kalomaze committed 1 year ago

Verified 238657db

flake.nix: fix for rocm 5.7 (#3853)

Tungsten842 committed 1 year ago

Verified 07178c98

ggml : move FP16 <-> FP32 code to ggml-impl.h (#3861)

ggerganov committed 1 year ago

Verified 207b5190

Extend llama_kv_cache_seq_rm to allow matching any sequence (#3843)

KerfuffleV2 committed 1 year ago

Verified 6e08281e

make : remove unnecessary dependency on build-info.h (#3842)

cebtenzzre committed 1 year ago

Verified 2046eb43

llama : fix kv shift bug (#3835)

ggerganov committed 1 year ago

Verified 71a09da3

ggml : quantization refactoring (#3833)

ggerganov committed 1 year ago

Verified d69d777c

flake : update flake.lock for newer transformers version + provide extra dev shell (#3797)

Green-Sky committed 1 year ago

Verified ff3bad83

metal : try cwd for ggml-metal.metal if bundle lookup fails (#3793)

akx committed 1 year ago

Verified 82a6646e

issues : change label from bug to bug-unconfirmed (#3748)

ggerganov committed 1 year ago

Verified ba231e8a

convert : ignore tokens if their IDs are within [0, vocab_size) (#3831)

ggerganov committed 1 year ago

Verified 8a2f2fea

llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747)

KerfuffleV2 committed 1 year ago

Verified bd6d9e20

llama : add option for greedy sampling with probs (#3813)

ggerganov committed 1 year ago

Verified ee1a0ec9

common : print that one line of the syntax help *also* to standard output (#3823)

HenkPoley committed 1 year ago

Verified 17746110

starcoder : add GPU offloading (#3827)

ggerganov committed 1 year ago

Verified fdee152e

speculative : ensure draft and target model vocab matches (#3812)

KerfuffleV2 committed 1 year ago

Verified 41aee4df

llama : correctly report GGUFv3 format (#3818)

cebtenzzre committed 1 year ago

Verified 6d459cbf

simple : fix batch handling (#3803)

tterrasson committed 1 year ago

Verified c8d6a1f3

cuda : improve text-generation and batched decoding performance (#3776)

ggerganov committed 1 year ago

Verified 2f9ec7e2

Older