You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As discussed with @cherryWangY, the enable_compression in TF DescrptDPA1Compat (directly inherited from DescrptSeAtten) doesn't consider the suffix of the type embedding. The type embedding in the original DescrptSeAtten doesn't contain a suffix, but that in DescrptDPA1Compat does.
Fix#4114 .
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
- **New Features**
- Enhanced compression capabilities in descriptor models with new
optional parameters for improved flexibility.
- Improved serialization processes for attention layers, allowing for
better handling of scaling factors and normalization.
- Dynamic tensor name construction in utility functions to accommodate
varying suffixes.
- **Bug Fixes**
- Adjusted method parameters to ensure compatibility and functionality
with new suffix options.
- **Tests**
- Introduced a new test suite to validate the functionality of the
TensorFlow-based descriptor model, ensuring consistent output with the
updated features.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
As discussed with @cherryWangY, the
enable_compression
in TFDescrptDPA1Compat
(directly inherited fromDescrptSeAtten
) doesn't consider the suffix of the type embedding. The type embedding in the originalDescrptSeAtten
doesn't contain a suffix, but that inDescrptDPA1Compat
does.deepmd-kit/deepmd/tf/descriptor/se_atten.py
Line 2223 in b6e27e4
Users may not touch this issue, as a descriptor usually has no suffix.
The text was updated successfully, but these errors were encountered: