songlab-cal / mogwai Goto Github PK
View Code? Open in Web Editor NEWLicense: BSD 3-Clause "New" or "Revised" License
mogwai's People
Contributors
Stargazers
Watchers
mogwai's Issues
mogwai-train doesn't track loss on wandb
All of the explicitly logged metrics such as auc, auc_apc, etc get logged on wandb. Loss is not logged on wandb, but if you explicitly try to log it via
self.log("loss", loss, on_step=True, on_epoch=False, prog_bar=True)it throws an error
File "/home/nickbhat/projects/mogwai/venv/bin/mogwai-train", line 33, in <module>
sys.exit(load_entry_point('mogwai-protein', 'console_scripts', 'mogwai-train')())
File "/home/nickbhat/projects/iclr-2021-factored-attention/mogwai/mogwai/train.py", line 101, in train
trainer.fit(model, msa_dm)
File "/home/nickbhat/projects/mogwai/venv/lib/python3.6/site-packages/pytorch_lightning/trainer/trainer.py", line 440, in fit
results = self.accelerator_backend.train()
File "/home/nickbhat/projects/mogwai/venv/lib/python3.6/site-packages/pytorch_lightning/accelerators/gpu_accelerator.py", line 54, in train
results = self.train_or_test()
File "/home/nickbhat/projects/mogwai/venv/lib/python3.6/site-packages/pytorch_lightning/accelerators/accelerator.py", line 66, in train_or_test
results = self.trainer.train()
File "/home/nickbhat/projects/mogwai/venv/lib/python3.6/site-packages/pytorch_lightning/trainer/trainer.py", line 483, in train
self.train_loop.run_training_epoch()
File "/home/nickbhat/projects/mogwai/venv/lib/python3.6/site-packages/pytorch_lightning/trainer/training_loop.py", line 550, in run_training_epoch
self.on_train_batch_end(epoch_output, epoch_end_outputs, batch, batch_idx, dataloader_idx)
File "/home/nickbhat/projects/mogwai/venv/lib/python3.6/site-packages/pytorch_lightning/trainer/training_loop.py", line 249, in on_train_batch_end
self.trainer.call_hook('on_train_batch_end', epoch_end_outputs, batch, batch_idx, dataloader_idx)
File "/home/nickbhat/projects/mogwai/venv/lib/python3.6/site-packages/pytorch_lightning/trainer/trainer.py", line 823, in call_hook
trainer_hook(*args, **kwargs)
File "/home/nickbhat/projects/mogwai/venv/lib/python3.6/site-packages/pytorch_lightning/trainer/callback_hook.py", line 147, in on_train_batch_end
callback.on_train_batch_end(self, self.get_model(), outputs, batch, batch_idx, dataloader_idx)
File "/home/nickbhat/projects/mogwai/venv/lib/python3.6/site-packages/pytorch_lightning/callbacks/progress.py", line 339, in on_train_batch_end
self.main_progress_bar.set_postfix(trainer.progress_bar_dict)
File "/home/nickbhat/projects/mogwai/venv/lib/python3.6/site-packages/pytorch_lightning/trainer/properties.py", line 155, in progress_bar_dict
return dict(**ref_model.get_progress_bar_dict(), **self.logger_connector.progress_bar_metrics)
TypeError: type object got multiple values for keyword argument 'loss'
Presumably this is because it's already logged, as loss is displayed in the progress bar.
Add Gremlin MLM
Training fails without structure
Currently the command
mogwai-train <msa_file.a3m>fails at
Traceback (most recent call last):
[...]
File "/home/nickbhat/projects/mogwai/mogwai/models/base_model.py", line 56, in training_step
auc = self.get_auc(do_apc=False)
File "/home/nickbhat/projects/mogwai/venv/lib/python3.6/site-packages/torch/autograd/grad_mode.py", line 15, in decorate_context
return func(*args, **kwargs)
File "/home/nickbhat/projects/mogwai/mogwai/models/base_model.py", line 111, in get_auc
"Model not provided with ground truth contacts, precision can't be computed."
ValueError: Model not provided with ground truth contacts, precision can't be computed.
This is because BaseModel always attempts to calculate auc in the training step,
Add figure logging to train.py
Log a few matplotlib images to wandb if wandb logging enabled.
Support TRRosetta NPZ Files
Necessary Changes
- Write a
TRRosetta_MSADatasetclass for parsing the MSA from the npz file. - Add a
.npzbranch toread_contacts.
confirm that the following does work.
mogwai-train <pdb.npz> --structure_file <pdb.npz>Single layer of attention `attention.py` uses incorrect regularization
Currently single layers of attention aren't training very well, afaik because we haven't finalized the regularization implementation
How to Represent Sequences?
We are currently one-hotting directly in the MSADataModules. We should do something smarter to be compatible with alternative tokenizations and amino acid embedding layers.
Open to thoughts here.
Deduplicate defaults
One hyperparameter has multiple defaults set all over the place, resulting in error-prone manual matching.
For example, num_repeats for RepeatDataset has a default in the add_args method of RepeatDataset and a default in the constructor of MSADataModule. This happens in all models as well.
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.