Joel Nicholls

thank you! Yes, sorry, I didn't write an update here yet. Essentially, a different error comes up after making this change. The error is basically not related to this one,...

The separate PR is over here https://github.com/blue-oil/blueoil/pull/1110 "fix loading of dataset class from yaml config" (still waiting for auto-tests to finish). I expect it will solve the checks here because...

@iizukak how do you think about these changes? One possible contentious point is that the `save_yaml` function remains, for saving the `meta.yaml`. I think that meta file may be important...

It's an important issue! Thank you for bringing it up. Yes, no seed is passed, so it will be generated by the system clock maybe? Another reason there is variability...

No problem! - I think the TensorFlow default is to use the system clock as seed. So, simply not setting a seed (or setting as `None`) will cause the system...

looks good ! nice and clean.

Yes, that's a good point, I agree about this, it would be good to specify the type of resampling. Otherwise, in the future, we might upgrade Pillow and not realize...

That's a good point, thank you for bringing it up. It looks like there are two places where `config.yaml` file is saved https://github.com/blue-oil/blueoil/blob/002f5408d11fd2a93457c5e7bc8dc876ace13ce0/blueoil/cmd/train.py#L299 and https://github.com/blue-oil/blueoil/blob/002f5408d11fd2a93457c5e7bc8dc876ace13ce0/blueoil/cmd/export.py#L146 In `train.py`, it happens immediately...

ah right, I should first request for owner approval, then after that, for readability approval. I'll remember for next time!

Yes, so I'm trying to show there is an error when loading the `config.yaml`. But if the `config.py` exists, then the `config.yaml` is not loaded, so the error doesn't happen....