conditional gen #34
Open
conditional gen #34
Conversation
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
…ltimeasurement noise
- Fixed noise level bug in denoiser_conditional - Cleaned up denoiser_conditional.py (now does not have multimeasurement) - Inserted a manual label override in parse_datasets_from_directory - New datasets and script for noise check experiment - Init graph recentering in ModelSamplingWrapper
…s and then swarms for a set of pdbs. Folder management is a bit janky right now and needs cleanup if you want to do it stepwise or if you want to add more swarms.
…till slow), sweep scripts
- Reorganized train scripts for enhanced sampling data, with a new script for running a single conditional model in train_enhanced_sampling_single.sh - Sample training script in configs/experiment/train_enhanced_position_conditioner.yaml
…ced denoising - Extended E3ConvConditional to accept additional input attributes - Added input_irrep_aggregator for combining node_attr with input_attr - Supports "3x1e" input attributes from spatiotemporal features - Maintains full E(3)-equivariance throughout processing pipeline - Subclass of Denoiser that extracts spatiotemporal features as input_attr - Modified xhat_normalized() to integrate spatiotemporal conditioning - Automatically extracts features from SpatioTemporalConditioner - Graceful fallback when spatiotemporal model unavailable - Updated to always return exactly 1 structure (not N_structures) - Improved documentation and error handling - Better integration with new input attribute system - src/jamun/hydra_config/model/arch/e3conv_conditional_with_input_attr.yaml - src/jamun/hydra_config/model/denoiser_conditional_with_input_attr.yaml - src/jamun/hydra_config/model/conditioners/spatiotemporal_with_input_attr.yaml - configs/experiment/train_enhanced_spatiotemporal_conditioner_with_input_attr.yaml - Fixed E3SpatioTemporal to use E3Conv instead of E3ConvConditional - Corrected import statements throughout spatiotemporal.py - Updated all irreps to use "3x1e" consistently for proper dimensionality - Fixed circular config references in YAML files - Corrected architecture call signatures (positional vs keyword args) - Fixed missing imports in average_squared_distance.py - Updated return type annotations for consistency - Improved edge addition logic in test scripts - Moved helper functions from scratch/transformer/helpers.py to appropriate utils modules - Enhanced test script with proper DenoiserWithInputAttr testing - Added comprehensive error handling and validation - Improved device handling and tensor shape management - All irreps configurations updated to "3x1e" for consistency - Factory functions properly configured for all architectures - Experiment configs optimized for training stability This implementation enables conditioning denoising models on rich spatiotemporal features while maintaining full equivariance and providing a clean, extensible architecture for future enhancements.
Spatiotemporal model has "ones" facility, for more elegantly turning off positional interactions The default positional attributes are now 3x0e--this should ensure full equivariance Spatiotemporal config has been broken down into standard and conditional Multimeasurement bug fix Sweep script for delta, friction sweep
1. New configs/slurm scripts for: high noise level, noise check experiment, and full swarm data. 2. Included equilibration mechanism in the baoab and aboba memory samplers. 3. A new wandb sweep script for sweeping over delta and friction.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
No description provided.