How to train and test ilql model? #48
-
|
I can train the bert model use bert.phase1, and test it without problem. |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
|
Please use |
Beta Was this translation helpful? Give feedback.
Please use
train.pyfor training.extract_policy.pyimplements the AWR algorithm, which is used in the original IQL paper. However,extract_policy.pyis for extracting a policy model from a trained Q-function model. If you simply want to make predictions, you can use the Q-function model directly, and I think there is no need to runextract_policy.py.