octo-models / octo

Octo is a transformer-based robot policy trained on a diverse mix of 800k robot trajectories.
https://octo-models.github.io/
MIT License
885 stars 166 forks source link

Save the original model config in train script #77

Open yushijinhun opened 7 months ago

yushijinhun commented 7 months ago

The train script removes the "oxe_kwargs" key in the config, and add "dataset_kwargs_list" & "sample_weights" to the config. Such behavior causes a TypeError: Object of type function is not JSON serializable(base) error when saving model checkpoint, because the standardize_fn we put into the model config is not serializable. This commit fixes the issue by preserving the unmodified config.