philschmid / deep-learning-pytorch-huggingface

MIT License
580 stars 138 forks source link

Clean up some typos; simplify some code; update some comments #53

Closed tomaarsen closed 1 month ago

tomaarsen commented 1 month ago

Hello!

Pull Request overview

Details

I ended up not changing the learning rate/batch size, etc., as I was unable to reproduce your findings exactly (due to the train_test method giving me different testing queries).

I also renamed "baseline" as we use it throughout the entire script, both for the baseline but also during training and after training. So I turned it into "financial" instead.

And I undid the column renaming: that's not necessary in Sentence Transformers. The columns are used in order, regardless of the names.

cc @philschmid