Arnaud Stiegler
Arnaud Stiegler
Happy to work on TrOCR (pytorch and TF)
Thank you! Let me know if there's anything I can help with :)
Oh yeah, you're right! Completely missed it, and it does solve the generation issue after 50 steps basically. ``` step: 0 train_loss: 8.3875150680542 prediction: ['
Good catch, just tried without the label smoothing and the losses now look much more normal: ``` step: 0 train_loss: 7.458827972412109 prediction: ['
Trying it right now! Will keep you updated once I got the results back :)
From my experiment, the training loss on larger datasets is indeed getting much lower (expected) but it doesn't seem to be solving the issue.
Losses overall look okay (with and without the label smoothing), but there seems to be some disconnect between the loss (both training and validation) value I'm getting and the actual...
Yeah, the model seems to be learning well on >3k images dataset with the change on the decoder config. This seems to be the root cause. Really good catch @gbarello-uipath...
Thanks for the answer! I didn't know about the predefined pipelines, not sure whether I missed them in the documentation. Are those just "random" pipelines or is there a specific...
One solution that works is: - `processor.tokenizer._tokenizer.pre_tokenizer.add_prefix_space = False` to prevent the model from using tokens preceded with a blank space - Add the token '1' to the tokenizer It's...