carpedm20 / carpedm20/ENAS-pytorch
a bug related to CNN search?
- Dominant language
- Python
- Stars
- 2.7k
- Forks
- 486
- PR merge metrics
- No merged PRs in 30d
Description
Hi, thanks for sharing the code. I am implementing the CNN part. I think the` block_idx` in `forward` function should be moded by 2 when CNN case as you used only two softmax. Could you check it? Thanks.
https://github.com/carpedm20/ENAS-pytorch/blob/25c4a89e17851d72e85566e16c81f0fd3749d58f/models/controller.py#L173-L177
Contributor guide
No contributing guide indexed for this repository
Research direction
Start with models/controller.py, especially the forward function at lines 173-177 linked in the issue, and trace the CNN case around block_idx and the two softmaxes. Confirm whether the index needs to be reduced modulo 2; done means CNN search uses the correct softmax index without changing other cases.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- machine-learning
- Issue type
- Bug
- Difficulty
- 2/5
- Estimated time
- 1-3 hours
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100