Hi, thanks, very welcome!

We used Python 3.10.0 for all original DNN runs, you can also check the exact list of packages in the logs, for example here.

What do you want to run in specific? If you want to deploy AutoXPCR on custom data, it is a bit tricky because you basically first need to predict which DNN to train, and then train the respective model. For our work we did not implement this as a "standalone" funtionality, and I currently don't have the time to add this.

But of course you are invited to look at the code and write a respective script on your own. It would need to follow these steps, let me know if you have issues:

Use the code from run_autoxpcr.py to train the meta-learners from the pre-assembled XPCR database
Use the code from run_autoxpcr.py to check how meta-features can be collected for custom data
Use the trained meta-learners to predict the performance of all DNNs based on your custom meta-features
Check the data_loader.py for information how your data should be loaded
Check the code from run.py to train and evaluate the best-estimated DNN on your custom data

Cheers!

Which Python version is recommended for this project? #1

Description

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions