carpedm20 / carpedm20/DCGAN-tensorflow
Always loads first checkpoint, won't read others.
- Dominant language
- JavaScript
- Stars
- 7.2k
- Forks
- 2.6k
- PR merge metrics
- No merged PRs in 30d
Description
Hey,
Thanks a lot for this model, I tried successfully to generate images at different resolutions and I am happy with it. While doing this I generated several checkpoints folder for different resolutions that are present in the _checkpoint_ folder.
However, when I try to load a former model and restore one of the checkpoints with this command:
`$ python main.py --dataset --input_height= --crop
`
It launches, load a checkpoint but produce samples from the first ever model I created, which is super low res, no matter the input_height or output_height I specify. I am pretty sure it keeps loading from the first ever checkpoint folder created.
I tried to delete the checkpoint that is launched but it won't recognise any other. I tried changing the name of the checkpoint folder I want to load to the one that is launched but it gives me this error:
`"InvalidArgumentError (see above for traceback): Restoring from checkpoint failed. This is most likely due to a mismatch between the current graph and the graph from the checkpoint. Please ensure that you have not altered the graph expected based on the checkpoint."`
In other words, I want to load a checkpoint that is in a folder images_64_512_512 but it keeps loading the folder image_64_64_64
I'm not sure how to change this, if you have any ideas or pointers, they are welcome.
Contributor guide
No contributing guide indexed for this repository
Assessment
This issue has not been assessed yet.