Closed evelynmitchell closed 8 months ago
Hello there, thank you for opening an Issue ! 🙏🏻 The team was notified and they will get back to you asap.
@evelynmitchell good catch, i don't know what I did to introduce this error. I tested it rigorously before pushing
You're welcome Kye
The second demo in the readme is the Ready to Train Model:
When I do a pip install from github in colab, and run this, it fails with a dimension mismatch, which, I believe is related to the ViT scaling of the original image (224) down to the floor of the patch size (224 // 16) and back up ^^2 The math is:
The error is that 196 is not 224.
I believe this would be fixed by using a text input length of 196.
README fix incoming.