9 forgetting pipeline #18
Draft
+3,150
−656
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
jack89roberts
added this to the Milestone 1: Working pipeline on small novel usecase milestone
jack89roberts
force-pushed
the
9-forgetting-pipeline
branch
2 times, most recently
from
7b122f7
to
2e186da
Compare
jack89roberts
force-pushed
the
9-forgetting-pipeline
branch
from
b8740ff
to
392151e
Compare
TOFU says it didn't work well
Steals a lot from fine-tuning branch - some placeholder forget configs and config classes - adds a data collator compatible with QAForgetDataset - adapts trainer loading etc. to have option of loading a forgetter instead BUG - Evaluation does not work, likely also needs adapting for expected data input format.
jack89roberts
force-pushed
the
9-forgetting-pipeline
branch
from
a877f2c
to
ad3522c
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Adds a
Forgetterclass for Gradient Ascent, Gradient Difference, KL, and "I don't know forgetters", which all inherit from the HuggingFaceTrainer, only modifyingcompute_loss.TODOs
Configs/Job Submission
All needs testing:
arcsf.config.experimenthas been adapted for the new config/job submission structure. It generates separate full, retain, and forget train scripts and configs.train.pyhas been modified to accept the new config structure and to work for both training and forgetting jobs, but not checked/tested at all yet.Forgetting Evaluation
eval_dataset/ any kind of evaluation during training as theevaluatefunction in the trainer class doesn't know what to when given 2 data inputs (forget / retain). Theevaluatefunction inarcsf.forget.baseneed to be implemented to use Jack D's evaluation code (and the trainer should be initialised witheval_datasetset to whatever our appropriate eval dataset instance is).