5 Commits

Author	SHA1	Message	Date
nielseni6	0efbfe7f4d	updated grad to create_graph during training, enabling better pgt_loss optimization	2024-10-22 23:40:46 -04:00
nielseni6	411157c18a	fixed yaml default	2024-10-21 13:05:25 -04:00
nielseni6	5953d3c9c6	changed save path to not override previous training	2024-10-21 13:04:57 -04:00
nielseni6	3a449d5a6c	Fixed PGT by including it in the loss function	2024-10-21 12:20:24 -04:00
nielseni6	38fa59edf2	PGT Training now functioning. Run run_pgt_train.py to train the model.	2024-10-16 19:26:41 -04:00