Seems this kind of mechanism of neural style transfer does not work for multi-gpu training.
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 9.3k
- Forks
- 4.4k
- Avg merge
- 1d 21h
- Merged PRs (30d)
- 4
Description
In neural style transfer tutorial, we just get the content layer and style layer as instances and put them in lists. Then in the main script, we do like this:
https://github.com/pytorch/tutorials/blob/beb7e27642808391bf31c97aae6d5ec0fc3a920f/advanced_source/neural_style_tutorial.py#L436-L439
It should not work when we perform multi-gpu training, although it dose not mean much for such task. However, for someone who want to add constraint on the intermediate layers of one model, they are likely to refer this kind of implementation, which is not satisfying.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with advanced_source/neural_style_tutorial.py around lines 436-439 and inspect how the content and style layer instances are collected and used. Reproduce or verify the behavior with multi-GPU training, then determine the appropriate correction for intermediate-layer constraints. Done should mean the tutorial mechanism works as intended across multiple GPUs, with behavior covered by an appropriate test or documented limitation.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- distributed-systems, machine-learning
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100