diff options
| author | Tim Dettmers <tim.dettmers@gmail.com> | 2022-07-25 14:02:14 -0700 |
|---|---|---|
| committer | Tim Dettmers <tim.dettmers@gmail.com> | 2022-07-25 14:02:14 -0700 |
| commit | 8b1fd32e3e4f5073fd055cb5f9261ec585f8cc2c (patch) | |
| tree | 76044424d73d02e1026c996b22b9da5061188387 /CHANGELOG.md | |
| parent | 7d2ecd30c044840ba5f161ec73e5eaf30ac8131d (diff) | |
Fixed makefile; fixed Ampere igemmlt_8 bug.
Diffstat (limited to 'CHANGELOG.md')
| -rw-r--r-- | CHANGELOG.md | 11 |
1 files changed, 11 insertions, 0 deletions
diff --git a/CHANGELOG.md b/CHANGELOG.md index fa20b15..08adfce 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -53,3 +53,14 @@ Bug fixes: Docs: - Added instructions how to solve "\_\_fatbinwrap_" errors. + + +### 0.30.0 + +#### 8-bit Inference Update + +Features: + - Added 8-bit matrix multiplication form cuBLAS, and cuBLASLt as well as multiple GEMM kernels (GEMM, GEMMEx, GEMMLt) + - Added 8-bit Linear layers with 8-bit Params that perform memory efficient inference with an option for 8-bit mixed precision matrix decomposition for inference without performance degradation + - Added quantization methods for "fake" quantization as well as optimized kernels vector-wise quantization and equalization as well as optimized cuBLASLt transformations + - CPU only build now available (Thank you, @mryab) |