Conversation
* Remove lightning from transcriptformer * Correct docstring * Make test device GPU * Remove device choice from config as flex attention is only GPU compatible * fixup! Remove device choice from config as flex attention is only GPU compatible
Change default emb mode to cell
* Stop helix-mrna from padding to max length and rather pad to max length of inputs * fixup! Stop helix-mrna from padding to max length and rather pad to max length of inputs
Attention maps
bputzeys
approved these changes
Jun 18, 2025
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This pull request introduces several updates to improve functionality, usability, and maintainability across multiple modules in the codebase. The most significant changes include adding support for attention outputs in embedding methods, improving installation instructions, and refining configuration and processing logic.
Enhancements to Embedding Methods:
Added an
output_attentionsparameter to embedding methods inGeneformerandscGPTmodels, enabling the return of attention maps for analysis. This includes updates to theget_embsfunction ingeneformer_utils.pyand_encodefunction inscgpt/model_dir/model.pyto handle attention outputs. [1] [2] [3] [4]Updated the
TransformerEncoderimport inscGPTto use a custom implementation for attention weight support.Installation and Documentation Improvements:
Improved installation instructions in
README.md, including a detailed guide for resolving issues withmamba-ssminstallation using.whlfiles.Updated the
evo_2/README.mdfile to use HTTPS for cloning thevortexrepository instead of SSH.Configuration and Defaults:
Changed the default embedding mode in
scGPTConfigfrom"cls"to"cell"for better usability. [1] [2]Removed the
precisionfield fromtranscriptformer_config.yamlto simplify configuration.Code Cleanup and Refinements:
Refactored error and warning messages in
Geneformerto improve readability and maintain consistency.Replaced
padding="max_length"withpadding="longest"in theprocess_datamethod of thehelix_mrnamodel for more efficient tokenization.Miscellaneous:
constants/hash_values.py.These changes collectively enhance the functionality, usability, and maintainability of the codebase, particularly in embedding workflows and installation processes.