Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

Knowledge Distillation as Semiparametric Inference

This code replicates the experiments of

Knowledge Distillation as Semiparametric Inference.
Tri Dao, Govinda M. Kamath, Vasilis Syrgkanis, and Lester Mackey.
International Conference on Learning Representations (ICLR). May 2021.

  title={Knowledge Distillation as Semiparametric Inference},
  author={Tri Dao and Govinda M Kamath and Vasilis Syrgkanis and Lester Mackey},
  booktitle={International Conference on Learning Representations},

Required packages

Python >= 3.7, Pytorch >= 1.7. More details in requirements.txt.

We use Pytorch-lightning to organize training code, Hydra to organize configurations.

[Optional] We use Ray for distributed training.

[Optional] We use Wandb for logging.

Code structure

├─ cfg               # Configuration files, for Hydra
├─ datamodules        # Code for datasets
├─ models            # Model implementations
├─ results           # json files containing raw results, and pdf files containing plots
├─ scripts           # Scripts to tune hyperparameters
├─  # Train student model, distilled from teacher model
├─            # Implementation of knowledge distillation losses
├─      # Distributed training with Ray [optional]
├─          # Train a model (e.g. teacher or student) from scratch
├─          # Utility functions


To train a ResNet18 on CIFAR10, and save model to disk:

python train.batch_size=512 model=resnet18 +save_checkpoint_path=checkpoints/resnet18/final.ckpt runner=pl

To train a CNN5 student:

python train.batch_size=512 train.kd.class=KDOrthoLoss train.gradient_clip_val=0.1 train.kd.temperature=2.0 runner=pl


This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit

When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact with any additional questions or comments.


This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.


Knowledge Distillation as Semiparametric Inference



Code of conduct

Security policy





No releases published


No packages published