diff options
Diffstat (limited to 'CHANGELOG.md')
| -rw-r--r-- | CHANGELOG.md | 19 |
1 files changed, 15 insertions, 4 deletions
diff --git a/CHANGELOG.md b/CHANGELOG.md index a5b29d8..0d3a379 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -1,11 +1,11 @@ -v0.0.21 +### 0.0.21 - Ampere, RTX 30 series GPUs now compatible with the library. -v0.0.22: +### 0.0.22: - Fixed an error where a `reset_parameters()` call on the `StableEmbedding` would lead to an error in older PyTorch versions (from 1.7.0). -v0.0.23: +### 0.0.23: Bugs: - Unified quantization API: each quantization function now returns `Q, S` where `Q` is the quantized tensor and `S` the quantization state which may hold absolute max values, a quantization map or more. For dequantization all functions now accept the inputs `Q, S` so that `Q` is dequantized with the quantization state `S`. @@ -18,7 +18,18 @@ Features: - Block-wise quantization routines now support CPU Tensors. -v0.0.24: +### 0.0.24: - Fixed a bug where a float/half conversion led to a compilation error for CUDA 11.1 on Turning GPUs. - removed Apex dependency for bnb LAMB + +### 0.0.25: + +Features: + - Added `skip_zeros` for block-wise and 32-bit optimizers. This ensures correct updates for sparse gradients and sparse models. + - Added support for Kepler GPUs. (#4) + +Bug fixes: + - fixed "undefined symbol: \_\_fatbinwrap_38" error for P100 GPUs on CUDA 10.1 (#5) + + |