Conflicting Info on Resolution

[michelleespranita]

Hi,

Thank you for open-sourcing the model weights for this project!

I’m seeing some conflicting info on resolutions and wanted to double-check:

1. Is the resolution for the ViT-L model 256 (as written in the GitHub README) or 224 (as written in the paper and config file)?
2. Is the resolution for the ViT-g models 384 (as written in the GitHub README) or 336 (as written in the config file)?
3. Are the published weights for the ViT-g models trained only on the MIMIC dataset or on the 18.1M proprietary echoes mentioned in the paper?

Thanks for your help and the great contribution to the community!