Recommended train_e2a configuration #2
Description
Hi, I'm trying to replicate the outcomes in your paper. However, with the example configuration in README.md, I couldn't achieve as good results as your demo page. Could you please share the train_e2a configuration that yielded your best results? Thank you!