Upload existing config #122
Replies: 2 comments
-
are you asking if you can load a previous training config? If so you could take the Training_congif.toml from a previous train with the settings you want. But you'll have to modify the save Arguments every time to match the name of your project with visual studio or another application that can read .toml files. I think you can also save the notebook to your drive and modify the code from there to augment the "default' settings as well. But as far as something native to the notebook, no there is no setting for that as far as I know. |
Beta Was this translation helpful? Give feedback.
-
There is a hidden feature for this. Before running the main trainer, create a new code cell and run this code: override_config_file = "/content/drive/MyDrive/config.toml"
override_dataset_config_file = "/content/drive/MyDrive/dataset_config.toml" Edit the locations as necessary. You can also choose only one or the other. Everything will be overriden and the values you choose in the colab will be ignored. This is not intuitive so I haven't included it in the colab. |
Beta Was this translation helpful? Give feedback.
-
Hi! Is is possible to upload old settings into the colab notebook?
Beta Was this translation helpful? Give feedback.
All reactions