Skip to content

Finetuning #255

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 21 commits into from
Apr 19, 2022
Merged

Finetuning #255

merged 21 commits into from
Apr 19, 2022

Conversation

@AkshitaB
Copy link
Contributor

@AkshitaB AkshitaB commented Apr 4, 2022 •
edited
Loading

Changes proposed in this pull request:

  • transformers::finetune step, which mostly mimics the TorchTrainStep with additions for tokenizing the data (and updating the model embeddings). It also contains model-specific defaults for the data collator.
  • RunGeneration now allows the trained model object as input.

Before submitting

  • I've read and followed all steps in the Making a pull request
    section of the CONTRIBUTING docs.
  • I've updated or added any relevant docstrings following the syntax described in the
    Writing docstrings section of the CONTRIBUTING docs.
  • If this PR fixes a bug, I've added a test that will fail without my fix.
  • If this PR adds a new feature, I've added tests that sufficiently cover my new functionality.

After submitting

  • All GitHub Actions jobs for my pull request have passed.

@AkshitaB AkshitaB marked this pull request as draft April 4, 2022 06:14
@AkshitaB AkshitaB requested review from dirkgr and epwalsh April 4, 2022 21:58
@AkshitaB AkshitaB marked this pull request as ready for review April 12, 2022 18:17
dirkgr
dirkgr requested changes Apr 12, 2022
View reviewed changes
examples/finetune/snli_steps.py


@Step.register("subset-data")
class SubsetData(Step):
Copy link
Member

@dirkgr dirkgr Apr 12, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We have the DatasetRemix step for Tango's DatasetDict. Can we have the same for HF's datasets?

Copy link
Contributor Author

@AkshitaB AkshitaB Apr 15, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Will work on it separately: #268 This is technically unrelated to finetuning.

tango/integrations/transformers/finetune.py

def run( # type: ignore[override]
self,
model: Lazy[Model],
Copy link
Member

@dirkgr dirkgr Apr 12, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If I want to do some sort of curriculum learning, can I pass in the output of another training step here?

Copy link
Contributor Author

@AkshitaB AkshitaB Apr 15, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This can be done once we fix this: #269

Copy link
Contributor Author

@AkshitaB AkshitaB Apr 15, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

And this: #270

@dirkgr dirkgr enabled auto-merge (squash) April 19, 2022 22:16
@dirkgr dirkgr merged commit 1083049 into main Apr 19, 2022
@dirkgr dirkgr deleted the finetuning branch April 19, 2022 22:23
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@dirkgr dirkgr dirkgr approved these changes

@epwalsh epwalsh Awaiting requested review from epwalsh

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@AkshitaB @dirkgr