TomLim

I am working on a similar WinoBias dataset to evaluate gender bias in coreference resolution. The prompts were already available in [promptsource](https://github.com/bigscience-workshop/promptsource/tree/86a51217b90bf0d3034cd05482b8dbb2f83aaec0/promptsource/templates/wino_bias).

The implementation is ready. I wrote a simple unittest to check the functionality of the tokenizer, but it makes two ci tests fail with the error, saying that tested `Tomlim/myt5-base`...

Sorry for late reply, I pushed the changes addressing the comments in the review. Can this PR be still re-opened? Thank you!

I fixed most of CLI errors, unfortunately the issue with run_tests still persitsts. It raises the error, saying that tested Tomlim/myt5-base is not a valid model handle, even though it...

I fixed the issue with loading the model by changing the `MyT5TokenizationTest` class, and all the added tests remained unchanged. I also added a sample "vocabulary" file to `fixtures` so...

Thank you from prompt reply! Please let me know if there are any other changes or tests that should be added before merging!

The conflict was caused by another model being added in the meantime. I resolved it and merged the new changes from the main 🙂.

@itazap @ArthurZucker can this PR be merged? I recently presented MyT5 at ACL and expect increased interest in the model (tokenizer) in the upcoming time. Thanks!

Thank you for the review, @ArthurZucker. I added a short description of the method in the model doc and added the missing "Copied from statements". To address the other points:...

Thank you! I merged the updates from the main to the branch and pushed an empty commit with `run-slow` as instructed by the bot 🤖. I see some tests are...