Age | Commit message (Expand) | Author |
2021-11-21 | Remove VQVAE stuff, did not work... | Gustaf Rydholm |
2021-11-17 | Update vqvae with new quantizer | Gustaf Rydholm |
2021-11-17 | Add new quantizer | Gustaf Rydholm |
2021-11-05 | Remove conv attention | Gustaf Rydholm |
2021-11-03 | Remove pixelcnn | Gustaf Rydholm |
2021-10-07 | Lint | Gustaf Rydholm |
2021-09-30 | Rename vqloss to commitment loss in vqvae network | Gustaf Rydholm |
2021-09-30 | Refactor residual block | Gustaf Rydholm |
2021-09-30 | Add num res blocks as a variable | Gustaf Rydholm |
2021-09-19 | Add ability to set use norm in vqvae | Gustaf Rydholm |
2021-09-18 | Add comment for quantization | Gustaf Rydholm |
2021-09-18 | Add more residual blocks in vqvae | Gustaf Rydholm |
2021-09-11 | Disable group norm in vqvae en/decoder | Gustaf Rydholm |
2021-09-11 | Rename vq loss to commitment loss | Gustaf Rydholm |
2021-08-08 | Add VQGAN and loss function | Gustaf Rydholm |
2021-08-06 | Fix VQVAE into en/decoder, bug in wandb artifact code uploading | Gustaf Rydholm |
2021-08-04 | Add conv attention, up and downsampling to vqvae module | Gustaf Rydholm |
2021-08-04 | Add char htr experiment, rename from ocr to htr, vqvae loss collapses | Gustaf Rydholm |
2021-08-04 | New VQVAE | Gustaf Rydholm |
2021-07-30 | attr bug fix, properly loading network | Gustaf Rydholm |
2021-04-25 | Efficient net and non working transformer model. | Gustaf Rydholm |
2021-04-22 | Fixed training script, able to train vqvae | Gustaf Rydholm |
2021-04-07 | Fixed typing and typos, train script load config, reformatted | Gustaf Rydholm |
2021-04-05 | Pre-commit fixes, optimizer loading fix | Gustaf Rydholm |
2021-03-20 | Inital commit for refactoring to lightning | Gustaf Rydholm |