ViTInpainter / README.md
aaawesome's picture
Epoch 1: improved checkpoint (val_loss=0.1466)
7f56ec4 verified
|
Raw
History Blame
543 Bytes
---
library_name: pytorch
tags:
- image-inpainting
- masked-image-modeling
- bidirectional-transformer
- custom-model
---
# Better Inpainting Transformer
This repository contains the improved inpainting model.
## Files
- `pytorch_model.bin`: inference weights
- `config.json`: config
- `training_metadata.json`: epoch and val loss for best uploaded checkpoint
- `model_last.pt`: resumable training checkpoint
- `model_best.pt`: best local checkpoint
## Notes
Recreate the `BetterInpaintTransformer` class and load the state dict manually.