A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit floating point (FP8) precision on Hopper and Ada GPUs, to provide better performance with lower memory utilization in both training and inference.
When the user calls import transformer_engine, we check for available frameworks in the top level __init__.py file. However, this is unnecessary. The user side code from, e.g. transformer_engine.pytorch import ... remains unchanged.
See #839 for details.
Fixes # (issue)
Type of change
[ ] Documentation change (change only to the documentation, either a fix or a new content)
[ ] Bug fix (non-breaking change which fixes an issue)
[x] New feature (non-breaking change which adds functionality)
[ ] Breaking change (fix or feature that would cause existing functionality to not work as expected)
Changes
Removes framework specific imports from top level init file.
Description
When the user calls import
transformer_engine, we check for available frameworks in the top level__init__.pyfile. However, this is unnecessary. The user side code from, e.g.transformer_engine.pytorch import ...remains unchanged. See #839 for details.Fixes # (issue)
Type of change
Changes
Removes framework specific imports from top level init file.
Checklist: