pytorch/examples

DCGAN default setting in cifar10 not working

Open

#263 opened on Dec 4, 2017

View on GitHub
 (2 comments) (1 reaction) (0 assignees)Python (21,634 stars) (9,429 forks)batch import
good first issue

Description

After run about 15 epoch, the generator seems to collapse to a parameter setting where it always emits the same point.

Did anyone run the script successfully?

Contributor guide

Tech stack
python
Domain
machine learning
Issue type
bug
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
half day
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
stale
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
needs investigation
Prerequisites
basic knowledge of GANsfamiliarity with PyTorchability to run Python scripts
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
20
Research direction
Examine the DCGAN implementation in the pytorch/examples repository, specifically the training loop for cifar10. Investigate whether the generator collapse is due to learning rate, discriminator strength, or mode collapse. Consider testing with different random seeds and monitoring generator loss. Look for any known fixes in related issues or pull requests.