Nexperia

This is the PyTorch implementation of the Nexperia image classification models.

Requirements

Python >= 3.6
PyTorch >= 1.0
CUDA
NumPy
pandas

Usage

Standard training

The main.py contains training and evaluation functions in standard training setting.

Runnable scripts

Training and evaluation using the default parameters

We provide our training scripts in directory scripts/. For a concrete example, we can use the command as below to train the default model (i.e., ResNet-34) on the Nexperia dataset:
```
$ bash scripts/nexperia/run_ce.sh [TRIAL_NAME]
```
The argument TRIAL_NAME is optional, it helps us to identify different trials of the same experiments without modifying the training script. The evaluation is automatically performed when training is finished.
Additional arguments include
- sat-es: initial epochs of SAT
- sat-alpha: the momentum term $\alpha$ of SAT
- mod: modification of SAT, e.g., bad_1, bad_boost
- eli: initial epochs of weighted CE for class i (from 1 to 10)
- ce-momentum: the momentum term of weighted CE
- arch: the architecture of backbone model, e.g., resnet34
- dataset: the dataset to train, e.g., nexperia_split, nexperia, CIFAR10

Reference

A report can be found in the report.

@inproceedings{kaiyihuang,
  title={The first progress report on },
  author={Huang, Lang and Zhang, Chao and Zhang, Hongyang},
  booktitle={Advances in Neural Information Processing Systems},
  volume={33},
  year={2020}
}

This is adapted from the paper.

@inproceedings{huang2020self,
  title={Self-Adaptive Training: beyond Empirical Risk Minimization},
  author={Huang, Lang and Zhang, Chao and Zhang, Hongyang},
  booktitle={Advances in Neural Information Processing Systems},
  volume={33},
  year={2020}
}

Contact

If you have any question about this code, feel free to open an issue or contact kaiyihuang@ust.hk.

corersky / nexperia