Numba parallel weights computation + dataloader #5
Conversation
This file contains 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
…_encoding function; minor linting, changed seq_name_to_sequence to a string instead of list of chars
…minor printouts to MSA_processing
passing z_dim into train_VAE script manually passing in args for VAE checkpoint reloading small typos in scripts
…to mixed batch joint training. Initialising the bias to mean(y_train) for much better convergence, still not great performance though. Moved parameter reading outside of main function so that we can override the z_dim size
saving vae checkpoints, checkpoint loading vs train from scratch, added sigmoid+bce loss, added 3 very long functions for mixed/alternating/frozen training modes to switch from command line, added linear model loss weight
… outputs are equal
…cord, will delete the bad ones
…gure out properly later
…apping file also added "identity" weights for completion
# Conflicts: # EVE/VAE_model.py # calc_weights.py # compute_evol_indices.py # data/mappings/example_mapping.csv # examples/Step0_optional_calc_weights.sh # examples/Step0_optional_calc_weights_slurm.sh # train_VAE.py # utils/data_utils.py # utils/weights.py
Added progress bar, weights-only calc mode
Fallback to normal mode also works well
…der, merged in changes from ProteinGym. Removed the aggregation methods for evol indices.
…default, tested with DLG4 (cherry picked from commit fcb7894)
loodvn
requested review from
aaronkollasch,
jonnyfrazer,
danieldritter and
pascalnotin
aaronkollasch
suggested changes
Mar 20, 2024
| @@ -11,7 +11,7 @@ EVE is a set of protein-specific models providing for any single amino acid muta | |||
| The end to end process to compute EVE scores consists of three consecutive steps: | |||
| 1. Train the Bayesian VAE on a re-weighted multiple sequence alignment (MSA) for the protein of interest => train_VAE.py | |||
| 2. Compute the evolutionary indices for all single amino acid mutations => compute_evol_indices.py | |||
| 3. Train a GMM to cluster variants on the basis of the evol indices then output scores and uncertainties on the class assignments => train_GMM_and_compute_EVE_scores.py | |||
| 3. Train a GMM to cluster variants on the basis of the qevol indices then output scores and uncertainties on the class assignments => train_GMM_and_compute_EVE_scores.py | |||
| def sample_latent(self, mu, log_var): | ||
| """ | ||
| Samples a latent vector via reparametrization trick | ||
| """ | ||
| eps = torch.randn_like(mu).to(self.device) | ||
| z = torch.exp(0.5*log_var) * eps + mu | ||
| z = torch.exp(0.5 * log_var) * eps + mu |
There was a problem hiding this comment.
it would be nice to keep linting/formatting to a separate PR so this one isn't as cluttered
|
|
||
| parser = argparse.ArgumentParser(description='Evol indices') | ||
| parser.add_argument('--MSA_data_folder', type=str, help='Folder where MSAs are stored') | ||
| parser.add_argument('--MSA_list', type=str, help='List of proteins and corresponding MSA file name') | ||
| parser.add_argument('--protein_index', type=int, help='Row index of protein in input mapping file') | ||
| parser.add_argument('--MSA_weights_location', type=str, help='Location where weights for each sequence in the MSA will be stored') | ||
| parser.add_argument('--theta_reweighting', type=float, help='Parameters for MSA sequence re-weighting') | ||
| # parser.add_argument('--MSA_weights_location', type=str, help='Location where weights for each sequence in the MSA will be stored') |
There was a problem hiding this comment.
should these arguments be deprecated instead of removed entirely?
| parser.add_argument('--training_logs_location', type=str, | ||
| help='Location of VAE model parameters') | ||
| parser.add_argument("--seed", type=int, help="Random seed", default=42) | ||
| parser.add_argument('--z_dim', type=int, help='Specify a different latent dim than in the params file') |
There was a problem hiding this comment.
this can be done by editing the params config file
| parser.add_argument('--force_load_weights', action='store_true', | ||
| help="Force loading of weights from MSA_weights_location (useful if you want to make sure you're using precalculated weights). Will fail if weight file doesn't exist.", | ||
| default=False) | ||
| parser.add_argument("--overwrite_weights", |
There was a problem hiding this comment.
is --overwrite_weights necessary?
| action="store_true", default=False) | ||
| parser.add_argument("--batch_size", type=int, | ||
| help="Batch size for training", default=None) | ||
| parser.add_argument("--experimental_stream_data", |
There was a problem hiding this comment.
we could consider making this the default and removing the CLI option, if it's judged robust enough
| print("Protein name: "+str(protein_name)) | ||
| print("MSA file: "+str(msa_location)) | ||
|
|
||
| if mapping_file["MSA_filename"].duplicated().any(): |
There was a problem hiding this comment.
for the new mapping colnames "MSA_filename" and "MSA_theta", we should probably accept the old filenames "protein_name" and "theta" for legacy mapping files
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
We should squash all the commit messages together, there's a lot of random scripts from other projects that were in (but I've now removed)