# yoshitomo-matsubara/torchdistill

A coding-free framework built on PyTorch for reproducible deep learning studies. PyTorch Ecosystem. 🏆26 knowledge distillation methods presented at TPAMI, CVPR, ICLR, ECCV, NeurIPS, ICCV, AAAI, etc are implemented so far. 🎁 Trained models, training logs and configurations are available for ensuring the reproducibiliy and benchmark.

Repository: https://github.com/yoshitomo-matsubara/torchdistill
Canonical: https://ross.abutalabs.com/products/torchdistill
Homepage: https://yoshitomo-matsubara.net/torchdistill/
Language: Python
License: MIT
License Family: permissive
Topics: knowledge-distillation, pytorch, image-classification, imagenet, object-detection, coco, semantic-segmentation, cifar10, cifar100, colab-notebook, google-colab, pascal-voc, nlp, natural-language-processing, transformer, glue, amazon-sagemaker-lab, pytorch-ecosystem, text-classification
Last push: 2026-08-20T07:05:20+00:00

## Health v2 (maintenance only)
Score: 86/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 98, release rhythm 64, longevity 100
- inputs: {"age_days": 2450, "days_push": 13, "days_rel": 28, "gap_med": 223, "n_releases_24m": 4}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1629, forks 145 (observed 2026-08-28T04:05:13.750482+00:00)

## What it is
torchdistill is a modular, configuration-driven PyTorch framework for knowledge distillation and general deep learning experiments, requiring no coding—experiments are defined via declarative YAML files. It implements 26+ state-of-the-art distillation methods and provides trained models, logs, and configs for reproducibility.

## Use cases
- distill a large teacher model into a smaller student model
- run knowledge distillation experiments without writing Python code
- reproduce published knowledge distillation results from papers
- extract intermediate representations from a model without changing its forward interface
- train image classification models on ImageNet or CIFAR via YAML config
- perform object detection or semantic segmentation experiments on COCO or PASCAL VOC
- run text classification experiments on GLUE tasks with Hugging Face models

## When to choose
- you need to apply or benchmark state-of-the-art knowledge distillation methods in PyTorch
- you want reproducible deep learning experiments driven by declarative YAML configs instead of code
- you need to distill models across vision and NLP tasks with pretrained checkpoints available
- you want to extract intermediate model features via forward hooks without modifying model code

## When to avoid
- you need a general-purpose training framework with no distillation focus and prefer writing plain PyTorch
- your project uses a framework other than PyTorch, such as TensorFlow or JAX
- you need highly custom training loops that don't fit the config-driven design
- you want a no-code GUI tool rather than YAML-based configuration

## Facets
- artifact type: framework
- maturity: active
- function: deep-learning, machine-learning, llm-training, nlp, image-processing, computer-vision, configuration-management
- domain: deep-learning, machine-learning, computer-vision, developer-tools
- platform: python, cross-platform
- tags: knowledge-distillation, pytorch, yaml-config, model-compression, teacher-student, reproducibility, image-classification, object-detection, semantic-segmentation, text-classification, natural-language-processing, gpu

## Member repositories
- yoshitomo-matsubara/torchdistill (main) score 86

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:13.750482+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:48:18.249770+00:00, confidence not recorded.
  - readme: https://github.com/yoshitomo-matsubara/torchdistill (fetched 2026-08-28T04:05:13.750482+00:00, sha 55215a0fe1bf)
  - homepage: https://yoshitomo-matsubara.net/torchdistill/ (fetched 2026-08-29T11:20:55.668942+00:00, sha 1ea9132f80e3)
  - registry_pypi: https://pypi.org/pypi/torchdistill/json (fetched 2026-08-29T11:20:55.679865+00:00, sha f776590e978e)
  - site_page: https://yoshitomo-matsubara.net/torchdistill/usage.html (fetched 2026-08-29T11:20:55.677972+00:00, sha 297ff07ebb57)
- Data as of 2026-08-30T08:39:29.467469+00:00.