ggerganov/llama.cpp

Pull Requests Commits
llama : make quantize example up to 2.7x faster (#3115)
cebtenzzre committed 2 years ago
Verified 98311c42
flake : allow $out/include to already exist (#3175)
jneem committed 2 years ago
Verified feea179e
cmake : compile ggml-rocm with -fpic when building shared library (#3158)
abetlen committed 2 years ago
Verified 769266a5
flake : include llama.h in nix output (#3159)
AsbjornOlling committed 2 years ago
Verified cf8238e7
make : fix clang++ detection, move some definitions to CPPFLAGS (#3155)
cebtenzzre committed 2 years ago
Verified 4b8560e7
CI: add FreeBSD & simplify CUDA windows (#3053)
alonfaraj committed 2 years ago
Verified 83a53b75
falcon : use stated vocab size (#2914)
akawrykow committed 2 years ago
Verified 5c872dbc
cmake : add relocatable Llama package (#2960)
mtmcp committed 2 years ago
Verified 990a5e22
docker : add gpu image CI builds (#3103)
canardleteer committed 2 years ago
Verified 980ab41a
gguf-py : support identity operation in TensorNameMap (#3095)
KerfuffleV2 committed 2 years ago
Verified e3940841
feature : support Baichuan serial models (#3009)
jameswu2014 committed 2 years ago
Verified 4c8643dd
speculative : add heuristic algorithm (#3006)
leng-yue committed 2 years ago
Verified 35f73049
whisper : tokenizer fix + re-enable tokenizer test for LLaMa (#3096)
goerch committed 2 years ago
Verified 71ca2fad
cmake : add a compiler flag check for FP16 format (#3086)
RossComputerGuy committed 2 years ago
Verified 1b6c650d
CUDA: mul_mat_q RDNA2 tunings (#2910)
JohannesGaessler committed 2 years ago
Verified 0a5eebb4
speculative: add --n-gpu-layers-draft option (#3063)
sozforex committed 2 years ago
Verified 84e72365
arm64 support for windows (#3007)
es0m committed 2 years ago
Verified b52b29ab
CUDA: fix LoRAs (#3130)
JohannesGaessler committed 2 years ago
Verified 4f7cd6ba
CUDA: fix mul_mat_q not used for output tensor (#3127)
JohannesGaessler committed 2 years ago
Verified 89e89599
CUDA: lower GPU latency + fix Windows performance (#3110)
JohannesGaessler committed 2 years ago
Verified d54a4027
cmake : support build for iOS/tvOS (#3116)
jhen0409 committed 2 years ago
Verified 1b0d0925
CUDA: add device number to error messages (#3112)
JohannesGaessler committed 2 years ago
Verified 8a4ca9af
metal : PP speedup (#3084)
ikawrakow committed 2 years ago
Verified f31b6f4e
convert: remove most of the n_mult usage in convert.py (#3098)
Green-Sky committed 2 years ago
Verified 6eeb4d90
metal : support for Swift (#3078)
kchro3 committed 2 years ago
Verified 21ac3a15
metal : support build for iOS/tvOS (#3089)
jhen0409 committed 2 years ago
Verified 4fd54779
flake : add train-text-from-scratch to flake.nix (#3042)
takov751 committed 2 years ago
Verified ec2a24fe
readme : fix typo (#3043)
eltociear committed 2 years ago
Verified 7d99aca7
metal : Q3_K speedup (#2995)
ikawrakow committed 2 years ago
Verified ba7ffbb2
examples : make n_ctx warning work again (#3066)
cebtenzzre committed 2 years ago
Verified e64f5b55
Terms Of Use Privacy Policy Refund Policy Impressum