]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 6986c78)
ggml : use F16 instead of F32 in Q4_0, Q4_1, Q8_0 (#1508)
authorGeorgi Gerganov <redacted>
Fri, 19 May 2023 19:17:18 +0000 (22:17 +0300)
committerGitHub <redacted>
Fri, 19 May 2023 19:17:18 +0000 (22:17 +0300)
commit2d5db48371052087a83974abda3767d1aedec598
treeca7e6ad4b2be21d96272aece6489b2f39c444ecbtree
parent6986c7835adc13ba3f9d933b95671bb1f3984dc6commit | diff
ggml : use F16 instead of F32 in Q4_0, Q4_1, Q8_0 (#1508)

* ggml : use F16 instead of F32 in Q4_0, Q4_1 and Q8_0

* llama : bump LLAMA_FILE_VERSION to 3

* cuda : update Q4 and Q8 dequantize kernels

* ggml : fix AVX dot products

* readme : update performance table + hot topics
README.md diff | blob | history
ggml-cuda.cu diff | blob | history
ggml.c diff | blob | history
ggml.h diff | blob | history
llama.cpp diff | blob | history
llama.h diff | blob | history
Packaging of ggml-org/llama.cpp