Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[Feature Request] Support dp convert-backend for models with type embedding for tf #4022

Closed
iProzd opened this issue Jul 25, 2024 · 1 comment · Fixed by #4320
Closed

[Feature Request] Support dp convert-backend for models with type embedding for tf #4022

iProzd opened this issue Jul 25, 2024 · 1 comment · Fixed by #4320
Assignees
@njzjz
Labels
enhancement
Milestone
v3.0.0

Comments

@iProzd
Copy link
Collaborator

iProzd commented Jul 25, 2024 •
edited
Loading

Summary

Support dp convert-backend for models with type embedding for tf.

Detailed Description

dp convert-backend now only supports models without type embedding for tf to convert to pt/dp.
Models with descriptors such as se_atten se_atten_v2 need to be added.

Also in issue #3997 .

Further Information, Files, and Links

No response
@iProzd iProzd mentioned this issue Jul 25, 2024
Merged
1 task
njzjz added a commit to njzjz/deepmd-kit that referenced this issue Jul 26, 2024
@iProzd @njzjz
fix(pt): fix get_dim for DescrptDPA1Compat (deepmodeling#4007)
f7aa626 
- [x] (Tomorrow) Test if it works for deepmodeling#3997. 

deepmodeling#3997 needs another fix in deepmodeling#4022 .

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit


- **New Features**
- Introduced a method to dynamically determine the output dimension of
the descriptor, enhancing its functionality and interaction with other
components.
- Improved tensor dimensionality handling in tests to ensure
compatibility with the new output dimension method.

<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Signed-off-by: Jinzhe Zeng <jinzhe.zeng@rutgers.edu>
Co-authored-by: Jinzhe Zeng <jinzhe.zeng@rutgers.edu>
@njzjz
Copy link
Member

njzjz commented Sep 16, 2024 •
edited
Loading

Should we add se_atten_v3 and make it in all backends fully compatible with each other?

mtaillefumier pushed a commit to mtaillefumier/deepmd-kit that referenced this issue Sep 18, 2024
@iProzd @njzjz
fix(pt): fix get_dim for DescrptDPA1Compat (deepmodeling#4007)
e9d3acb 
- [x] (Tomorrow) Test if it works for deepmodeling#3997. 

deepmodeling#3997 needs another fix in deepmodeling#4022 .

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit


- **New Features**
- Introduced a method to dynamically determine the output dimension of
the descriptor, enhancing its functionality and interaction with other
components.
- Improved tensor dimensionality handling in tests to ensure
compatibility with the new output dimension method.

<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Signed-off-by: Jinzhe Zeng <jinzhe.zeng@rutgers.edu>
Co-authored-by: Jinzhe Zeng <jinzhe.zeng@rutgers.edu>
@iProzd iProzd moved this to enhance in DeePMD-kit V3.0.0 RC Sep 26, 2024
@njzjz njzjz added this to the v3.0.0 milestone Sep 26, 2024
@njzjz njzjz modified the milestones: v3.0.0, v3.1.0 Oct 28, 2024
@njzjz njzjz self-assigned this Nov 7, 2024
@njzjz njzjz modified the milestones: v3.1.0, v3.0.0 Nov 7, 2024
njzjz added a commit to njzjz/deepmd-kit that referenced this issue Nov 7, 2024
@njzjz
fix: consistent DPA-1 model
4de5511 
Fix deepmodeling#4022.
Note that `smooth_type_embedding==True` is not consistent between TF and others.
Also fix several issues.

Signed-off-by: Jinzhe Zeng <jinzhe.zeng@rutgers.edu>
@njzjz njzjz mentioned this issue Nov 7, 2024
Merged
@njzjz njzjz linked a pull request Nov 7, 2024 that will close this issue
fix: consistent DPA-1 model #4320
Merged
github-merge-queue bot pushed a commit that referenced this issue Nov 8, 2024
@njzjz
fix: consistent DPA-1 model (#4320)
0199ad5 
Fix #4022.
Note that `smooth_type_embedding==True` is not consistent between TF and
others.
Also, fix several issues.

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

## Release Notes

- **New Features**
- Enhanced configurability of descriptors with new optional parameters
for type mapping and type count.
- Introduction of a new class `DescrptSeAttenV2` for advanced attention
mechanisms.
- Added a new unit test framework for validating energy models across
multiple backends.

- **Bug Fixes**
- Improved error handling in descriptor serialization methods to prevent
unsupported operations.

- **Documentation**
- Updated backend documentation to include JAX support and clarify file
extensions for various backends.

- **Style**
	- Enhanced readability of error messages in fitting classes.

- **Tests**
- Comprehensive unit tests added for energy models across different
machine learning frameworks.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Signed-off-by: Jinzhe Zeng <jinzhe.zeng@rutgers.edu>
@njzjz njzjz closed this as completed Nov 8, 2024
@github-project-automation github-project-automation bot moved this from enhance to Done in DeePMD-kit V3.0.0 RC Nov 8, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@njzjz njzjz

Labels
enhancement
Projects
Multiple backend support for DeePMD-kit
Archived in project
DeePMD-kit V3.0.0 RC
Status: Done
Milestone
v3.0.0
Development

Successfully merging a pull request may close this issue.

fix: consistent DPA-1 model njzjz/deepmd-kit
2 participants
@njzjz @iProzd