Lightning-AI/pytorch-lightning

An example for knowledge distillation. Especially for the load the teacher model's weights , and train the student model

Open

#17,592 opened on May 8, 2023

View on GitHub
 (2 comments) (1 reaction) (0 assignees)Python (3,233 forks)batch import
examplefeaturehelp wanted

Repository metrics

Stars
 (26,687 stars)
PR merge metrics
 (Avg merge 9d 15h) (3 merged PRs in 30d)

Description

Description & Motivation

No response

Pitch

A example for knowledge distillation. Especially for the load the teacher model's weight ,and train the student model.

Now I have a trained teacher_model. When I want to train the student model, I need to load the teacher model's weights and set the teacher model to eval().

I am confused about the specific steps to do above using the Lightning. Could I get an example for it?

Alternatives

An example for knowledge distillation. Especially for the load the teacher model's weight ,and train the student model

Additional context

An example for knowledge distillation. Especially for the load the teacher model's weight ,and train the student model.

cc @borda

Contributor guide

Research direction
Look for existing examples of knowledge distillation in the Lightning repository or provide guidance on instantiating a teacher model, loading its pretrained weights, setting it to eval mode, and defining a student model. Implement a LightningModule that handles distillation loss and training logic.
Tech stack
pythonpytorch
Domain
machine learningai
Issue type
Documentation
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
2
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-3 hours
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Clear
Prerequisites
PythonPyTorch
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
70