Uh oh!
There was an error while loading. Please reload this page.
print selene_sdk version, add config and model file to output, add ra… - #191
print selene_sdk version, add config and model file to output, add ra…#191ygliu2016 wants to merge 2 commits into
Conversation
…ndom suffix to output directory name
kathyxchen
commented
Dec 20, 2022
Thanks Aaron for adding me to this PR, and thank you @ygliu2016 for your work on this! I've left some comments--feel free to tag me in a comment once you've updated / tested the changes based on the review, and let me know if you have any questions in the meantime. Overall it looks good! Thanks again :) |
ygliu2016
commented
Dec 20, 2022
H, Kathy, thanks for coming back. Would you please tell me where are your comments? In which repository? Thanks. |
kathyxchen
commented
Dec 20, 2022
The comments are directly in the changes you made - so if you scroll up above my comment or click 'files changed' tab you should be able to see what I wrote at the corresponding lines of code! |
ygliu2016
commented
Jan 10, 2023
I am so sorry. But I can not see you comments at repository ygliu2016/selene, branch info. |
Uh oh!
There was an error while loading. Please reload this page.
kathyxchen
commented
Jan 10, 2023
Hi @ygliu2016 , were you able to find the comments after you made the test comment? It's at https://github.com/FunctionLab/selene/pull/191/files "Files Changed" tab |
ygliu2016
commented
Jan 11, 2023
No, I cannot. Unfortunately, I can only see my test comment. Can you comment directly in the code? |
kathyxchen
left a comment
There was a problem hiding this comment.
Hi @ygliu2016, I've resubmitted the comments, hopefully this time it works!! Thanks for your patience :)
Uh oh!
There was an error while loading. Please reload this page.
| operations = configs["ops"] | ||
| #print selene_sdk version | ||
| if "selene_sdk_version" not in configs: |
There was a problem hiding this comment.
maybe simplify this to version. I think it's also OK to assume that version is never specified in the config and to add it automatically, e.g. moving from selene_sdk import version to the top of the page. I think you can also directly do import selene_sdk and use selene_sdk.__version__ -- not sure which is better practice.
There was a problem hiding this comment.
Not sure about name "selene_sdk_version" vs "version". May want to keep the longer version at this time.
There was a problem hiding this comment.
OK I thought about it more, and I think it's better to have this be automatically populated? Let's just remove lines 322 and 323 outright.
There was a problem hiding this comment.
Please adjust based on comments above, I think it makes sense to just auto-populate since almost no one will try to specify it manually in their config
Uh oh!
There was an error while loading. Please reload this page.
| np.random.seed(seed) | ||
| torch.manual_seed(seed) | ||
| torch.cuda.manual_seed_all(seed) | ||
| #torch.backends.cudnn.deterministic = True |
There was a problem hiding this comment.
why are these commented out?
There was a problem hiding this comment.
It is your code, which is not included in the main repository of selene. I am not sure if the code is necessary and thus copied here but commented them. It seems to me that with a seed, there is already deterministic in it.
There was a problem hiding this comment.
Thanks for the clarification. Please un-comment this as we need it for ensuring all aspects are deterministic.
There was a problem hiding this comment.
Same with the commented out line in 365, if that is my code.
There was a problem hiding this comment.
can you remove the commented out lines in 364 and 365?
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
ygliu2016
commented
Jan 11, 2023
Thank you very much for the comments and in fact, for your patience as well. |
kathyxchen
left a comment
There was a problem hiding this comment.
Added last few comments! Let me know when these final changes have been incorporated and tested. :) Thank you
| operations = configs["ops"] | ||
| #print selene_sdk version | ||
| if "selene_sdk_version" not in configs: |
There was a problem hiding this comment.
OK I thought about it more, and I think it's better to have this be automatically populated? Let's just remove lines 322 and 323 outright.
| np.random.seed(seed) | ||
| torch.manual_seed(seed) | ||
| torch.cuda.manual_seed_all(seed) | ||
| #torch.backends.cudnn.deterministic = True |
There was a problem hiding this comment.
Thanks for the clarification. Please un-comment this as we need it for ensuring all aspects are deterministic.
| np.random.seed(seed) | ||
| torch.manual_seed(seed) | ||
| torch.cuda.manual_seed_all(seed) | ||
| #torch.backends.cudnn.deterministic = True |
There was a problem hiding this comment.
Same with the commented out line in 365, if that is my code.
Uh oh!
There was an error while loading. Please reload this page.
jzthree
commented
Jan 21, 2023
I think |
Oh thanks - that’s good to know! @jzthree |
kathyxchen
left a comment
There was a problem hiding this comment.
hi @ygliu2016, I made another pass - some of the conversations you marked resolved I had some minor comments to fully resolve them. Can you tag me when you've finished the changes so I know to review this PR again? thanks!
| random_seed: 1447 | ||
| output_dir: ./training_outputs | ||
| create_subdirectory: False | ||
| create_subdirectory: True |
There was a problem hiding this comment.
I assume this was changed because of testing - can this be changed back since it's not relevant to the PR?
| # copy model file or directory to output | ||
| model_input = configs['model']['path'] | ||
| if os.path.isdir(model_input): # copy the directory | ||
| shutil.copytree (model_input, |
There was a problem hiding this comment.
spacing here has not been fixed still. Please adjust so it adheres to the current style in the codebase, thank you!
| configs = load_path(configs_file, instantiate=False) | ||
| operations = configs["ops"] | ||
| #print selene_sdk version |
There was a problem hiding this comment.
remove commented out print statement
| operations = configs["ops"] | ||
| #print selene_sdk version | ||
| if "selene_sdk_version" not in configs: |
There was a problem hiding this comment.
Please adjust based on comments above, I think it makes sense to just auto-populate since almost no one will try to specify it manually in their config
| np.random.seed(seed) | ||
| torch.manual_seed(seed) | ||
| torch.cuda.manual_seed_all(seed) | ||
| #torch.backends.cudnn.deterministic = True |
There was a problem hiding this comment.
can you remove the commented out lines in 364 and 365?
…ndom suffix to output directory name
Reference Issues/PRs
What does this implement/fix? Explain your changes.
What testing did you do to verify the changes in this PR?