Good news! We provided a colab playground where you can build your own Predecessors Combination and see its perforamnce!
A PyTorch implementation for Calibrating a Deep Neural Network with Its Predecessors. Licensed under the Apache License, Version 2.0.
This project is based on Focal Calibration.
Python >= 3.7.5, PyTorch == 1.3.1, torchvision == 0.4.2
train_search.py: search for architectures;
python train_search.py --epochs=350 --seed=1 --warm_up_population=100 --gumbel_scale=1e-1 --arch_learning_rate=1e-2 --memory_size=100 --predictor_warm_up=500 --dataset_name=cifar10 --model_name=resnet50 --ftlr=1e-3
train.py: train models;
python train.py --model resnet50 --dataset-root DATASET_DIR --loss cross_entropy
test_combination.py: test predecessor combinations;
python test_combination.py --dataset cifar10 --model resnet50 --tune_epoch 1 --combination 186,313,299,139,189 -log --weight_folder WEIGHT_DIR --lr=1e-4
If you are using the code/data in this repo, please consider citing:
@misc{tao2023calibrating,
title={Calibrating a Deep Neural Network with Its Predecessors},
author={Linwei Tao and Minjing Dong and Daochang Liu and Changming Sun and Chang Xu},
year={2023},
eprint={2302.06245},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent