trainML notebooks can now be forked into new instances to enable easy parallel experimentation. Unlike other cloud notebooks, when you fork a trainML notebook, the entire working directory is copied. All datasets, checkpoints, and other data are copied into the new notebook.
How It Works
From the Notebook Dashboard, select the notebook you wish to fork and click the
Copy button from the dashboard menu. The
Copy button is only enabled when a single notebook is selected and that notebook is either
stopped. To copy the entire notebook with its data, select the
Full copy type. You can modify the name and resources for the copy as needed. For example, if you are developing your model on a more affordable GPU type, but need to get the results of your experiment done faster, you can select a larger GPU type during the copy process. The available working directory space can also be increased if desired, but not decreased.
Copy button on the form to initiate the copy. Once the copy is complete, the new notebook will automatically start. The copied notebook is now fully independent of the original, so any changes to it will not affect the original and vice versa.
If you would like to create a new notebook with the same configuration as the existing notebook without copying any of the data, you can select the
Configuration Only option as the copy type. This option has the same effect as using the
Create button and having all the values preset to the values of the current notebook.