Jennifer Wei
c715b13830
Merge remote-tracking branch 'refs/remotes/jnwei/pl_upgrades' into pl_upgrades
2024-05-06 08:42:37 +00:00
Jennifer
0a8ae6a72c
Updates low_precision check to use current precision settings.
2024-05-06 08:32:04 +00:00
Jennifer
19c815888d
adds reload_dataloaders_every_n_epochs flag
2024-05-06 08:30:34 +00:00
Jennifer
3eed6cb8be
more logging changes
2024-05-06 08:29:30 +00:00
Jennifer
53cdb247ab
add metric logging to progress bar.
2024-05-06 08:28:40 +00:00
Jennifer
7de0ab00db
first pass changes to run with pl 2.1
2024-05-06 08:16:08 +00:00
Jennifer
142c5e65f1
Updates organization of command line flags for pl.Trainer
2024-05-02 03:44:03 -04:00
Jennifer
3a4ed7534d
Removes OF copy of zero_to_fp32.py favoring deepspeed.util version
2024-05-02 03:22:55 -04:00
Jennifer
e4f9af23e7
Merge branch 'setup-improvements' of https://github.com/aqlaboratory/openfold into setup-improvements
2024-05-02 03:11:15 -04:00
Jennifer Wei
1952582615
changes required for pytorch2
2024-05-02 02:54:34 -04:00
Jennifer
226c7bac08
DNS: Local profiler changes
2024-05-02 02:54:34 -04:00
Jennifer
eba00d85f7
adds config recording to wandb
2024-05-02 02:54:06 -04:00
Jennifer
870e0db32b
Adds experiment_config_json for setting custom configurations with a json.
2024-05-02 02:54:06 -04:00
Jennifer
67454c3982
bugfixes and adds a section to convert optim files
2024-05-02 02:52:04 -04:00
Jennifer Wei
60fbe53990
changes required for pytorch2
2024-05-02 02:36:13 -04:00
Jennifer
ec6058cd56
DNS: Local profiler changes
2024-05-02 02:36:12 -04:00
Jennifer
6b77062dd9
adds config recording to wandb
2024-05-02 02:36:12 -04:00
Jennifer
0f2b292913
Adds experiment_config_json for setting custom configurations with a json.
2024-05-02 02:36:12 -04:00
Jennifer
5bfad07412
Changes resume_model_weights_only flag in train_openfold to reload weights only, without parsing a time step.
2024-05-02 02:36:12 -04:00
Jennifer
a7216c73cf
changes reloading weights to account for possible state_dict headers.
2024-05-02 02:36:12 -04:00
Jennifer
711ba0abe2
Add log statement to weight conversion script
2024-05-02 02:36:12 -04:00
Jennifer
dd1dd64137
bugfixes and adds a section to convert optim files
2024-05-02 02:36:12 -04:00
Jennifer
ea142a0a68
fixes deepspeed function definition.
2024-04-19 06:40:37 -04:00
Jennifer
1ae833bfc8
Updates low_precision check to use current precision settings.
2024-04-19 06:14:08 -04:00
Jennifer
bbf989a7bf
psivant local pl2 upgrades without mpi configuration
2024-04-16 20:48:33 -04:00
Jennifer Wei
ce000c6061
changes required for pytorch2
2024-04-12 10:14:57 +00:00
Jennifer
523adaf448
adds reload_dataloaders_every_n_epochs flag
2024-04-11 02:57:49 -04:00
Jennifer
577219c112
Removes OF copy of zero_to_fp32.py favoring deepspeed.util version
2024-04-02 05:01:33 -04:00
Jennifer
862635834a
add paren to save_hyperparameters
2024-04-02 02:49:16 -04:00
Jennifer
5ff5177bc6
more logging changes
2024-04-01 04:36:34 -04:00
Jennifer
d8418293cd
DNS: Local profiler changes
2024-03-29 04:28:27 -04:00
Jennifer
0c3435cc75
add metric logging to progress bar.
2024-03-27 06:06:05 -04:00
Jennifer
cfd2e71981
seed workers fix and validation_epoch_end extra argument
2024-03-25 04:39:39 -04:00
Lukas Jarosch
a56ea9b564
Fix distributed seeding behavior
...
This adds workers=True to the Lightning seed_everything function which guarantees different random states across all processes in distributed training. Prior to that some processes on different GPUs with the same worker ID could share the same random state.
Note that this will break reproducibility between runs prior to and after this change.
Also removes the seed and supress_output modules that were not used anymore in OpenFold.
2024-03-19 10:05:22 -07:00
Jennifer
6dc34d71f5
first pass changes to run with pl 2.1
2024-03-19 05:30:07 -04:00
Jennifer
6ba0a594c9
adds config recording to wandb
2024-03-13 01:53:37 -04:00
Jennifer
d1fe024b20
Adds experiment_config_json for setting custom configurations with a json.
2024-03-13 01:22:46 -04:00
Jennifer
f4df72173d
Changes resume_model_weights_only flag in train_openfold to reload weights only, without parsing a time step.
2024-02-22 04:03:06 -05:00
Jennifer
ddfccd56d7
changes reloading weights to account for possible state_dict headers.
2024-02-20 04:40:40 -05:00
Jennifer
e31e03781f
Add log statement to weight conversion script
2024-02-14 00:48:00 -05:00
Jennifer
775f77dd12
bugfixes and adds a section to convert optim files
2024-02-12 05:14:53 -05:00
Christina Floristean
6275091c96
Fixed learning rate scheduler issue, returned to original msa file parsing
2024-02-07 15:47:24 -05:00
Christina Floristean
f65b75fe48
Fix for loading old OF weights into refactored model
2023-11-03 14:26:18 -04:00
Christina Floristean
d7c115371e
Merge branch 'main' into multimer
2023-10-27 12:09:39 -04:00
Christina Floristean
7c7dffd081
Refactoring multimer perm alignment, temporary fixes before removing recycling dimension
2023-10-16 21:20:38 -04:00
Christina Floristean
0cf1541ce1
Refactoring multimer data pipeline and permutation alignment.
2023-10-16 15:40:40 -04:00
Sachin Kadyan
6012b9e1c1
Default value for --use_single_seq_mode arg
2023-10-10 16:33:07 -04:00
Sachin Kadyan
395a9f1ba8
Added flag in training script for using sequence embeddings
2023-10-10 16:33:07 -04:00
Geoffrey Yu
3aaf0ca89f
update trainining code with new input from new multimer pipeline
2023-09-25 15:02:12 +02:00
Dingquan Yu
da92663d80
Update train_openfold.py
2023-09-10 20:10:16 +02:00
