diff options
| author | Gustaf Rydholm <gustaf.rydholm@gmail.com> | 2021-08-04 15:27:48 +0200 |
|---|---|---|
| committer | Gustaf Rydholm <gustaf.rydholm@gmail.com> | 2021-08-04 15:27:48 +0200 |
| commit | efe850821b88306481ab5aa2a5f79a2581e4458c (patch) | |
| tree | 6587de97cc8cbe3c427915f81de472b5d2bab3cb | |
| parent | 04c40f790e405ced6e6b90cf0a8aea268b9345c4 (diff) | |
Updates readme
| -rw-r--r-- | README.md | 9 |
1 files changed, 6 insertions, 3 deletions
@@ -10,6 +10,7 @@ TBC ### Build word piece dataset Extract text from the iam dataset: +TODO: Fix these! ``` python extract-iam-text --use_words --save_text train.txt --save_tokens letters.txt ``` @@ -27,15 +28,17 @@ python build-transitions --tokens iamdb_1kwp_tokens_1000.txt --lexicon iamdb_1kw ## Todo - [x] Efficient-net b0 + transformer decoder -- [ ] Load everything with hydra, get it to work -- [ ] Train network +- [x] Load everything with hydra, get it to work +- [x] Train network +- [ ] Get VQVAE2 to work and not get loss NAN +- [ ] Local attention for target sequence +- [ ] Rotary embedding for target sequence - [ ] Tests - [ ] Evaluation - [ ] Wandb artifact fetcher - [ ] attr refactor - [ ] Refactor once more - [ ] fix linting -- [ ] fix loading of transform iam paragraph ## Run Sweeps |