[dattri.benchmark] Add CIFAR-2 and MAESTRO support in `dattri/scripts/dattri_retrain.py` - Githubissues

TRAIS-Lab / dattri

`dattri` is a PyTorch library for developing, benchmarking, and deploying efficient data attribution algorithms.

https://trais-lab.github.io/dattri/

24 stars 8 forks source link

[dattri.benchmark] Add CIFAR-2 and MAESTRO support in `dattri/scripts/dattri_retrain.py` #77

Closed tingwl0122 closed 3 months ago

tingwl0122 commented 3 months ago

Description

1. Motivation and Context

This PR merges cifar2-resnet9 and maestro-musictransformer settings into the current retrain script.

2. Summary of the change

add cifar2_resnet9/maestro_musictransformer support in retrain script
slightly modify Maestro benchmark files
modify gitignore to ignore maestro files and slurm files
[x] TODO: check MAESTRO functionality after subset sampler in #66 is merged
[x] TODO: make device work in run_model

3. What tests have been added/updated for the change?

[x] N/A: No test will be added (please justify)

tingwl0122 commented 3 months ago

others LGTM

will merge this after a small-scale test using the sampler

tingwl0122 commented 3 months ago

others LGTM

will merge this after a small-scale test using the sampler

With the subset sampler, I have successfully utilized the retrain script to get normal results. @TheaperDeng, I guess this is ready to merge.