Blake Chi

Hi @lewtun, Just want to know is this issue in progress now? If not, I have some bandwidth to work on it. Or maybe we should wait until #69 merged?...

Hey @lewtun, Great! And the API looks good to me. But just a suggestion. Since `trainer.freeze` is kind of saying we're going to freeze all of the model (body +...

Hi @lewtun, good to hear that! I think I will begin with implementing a new class `SetFitHead` and then integrate it into `SetFitModel` by adding one more argument: `use_differentiable_head`. Does...

Ok, I will start to implement it!

Hi @lewtun, I think I'm in the half way (or more) of the implementation and a question raised in my mind. From the API you suggested as below, after we...

Sounds great! Just added that. Start to work on testing.

Hi @lewtun, Just opened a pull request for this issue! Please have a check. I haven't trained using the differentiable head, could you suggest a script for training? Thanks!

> Absolutely incredible work on implementing a pure `torch` model @blakechi 🔥 ! > > I've left a few comments / questions, but overall this is looking really good. >...

Here is the snippet for using the differentiable head (partially copied from `README.md`): ```python from datasets import load_dataset from sentence_transformers.losses import CosineSimilarityLoss from setfit import SetFitModel, SetFitTrainer # Load a...

Hi @lewtun, Sorry for the late. I was packed by other stuff. Ya, that sounds great to me. :) - Sure, just resolved it! - Thanks! I also ran experiments...