By mistake I submitted the same code 2 times because due to some error and ran the submission #2 so could you please remove the submission one.
can we edit the file after uploading it.
It is fine. You can submit up to 15 submissions per day.
No once a submission has been submitted, no further modification is allowed.
Can you help us to resolve this issue?
How we can specify the version in .ipynb file so that requirement_original.txt install correct version.
You can specify the versions by using version specifiers.
However your error look weird, could you give me your Run ID?
my run id is 29820.
Can you please reply a bit early?
Sorry, but the problem is as simple as just re-ordering packages.
It seems that torch needs to be installed before torchsort: Cannot install using poetry · Issue #58 · teddykoker/torchsort · GitHub
xformers have the same problem: no model named 'torch' · Issue #740 · facebookresearch/xformers · GitHub
And as you can see, no one like that:
To fix the problem, I need to update the runner to look for “primary” libraries like torch, which need to be installed in a separate command before the rest of requirements.txt.
And this kind of thing takes time to make sure the issues are resolved without affecting other users.
In the meantime, try not to use torchsort or use an alternative.
I will notify you when the bug is fixed.
If you want to get answered faster, please consider joining the Discord server.
Thank you, @enzo
I fixed the issue.
Could you create a new run please?
Dear Enzo, thank you for working with us and being patient. So the issue now seems that the train and infer function calls have been commented out. How to pass the data paths to the train/infer functions? We are also attaching the screenshots for your reference.
Run ID #29903, please.
Please review
In the train function, you are supposed to os.listdir(data_directory_path) and loop however you see fit.
And that mean that the data_file_path is not available in the train function:
As for why the code is commented out, when you submit a notebook, everything that is not a class, a function, or an import is automatically commented out.
The runner will call the train and infer functions for you.
This is to avoid floating code that would be executed during the loading of the main.py file.
Dear enzo,
I am experiencing an issue while trying to submit my results for the competition. Below is the error message I see in the logs:
You final model weight 17.80gb…
Unfortunately, our infrastructure rejected it without giving a proper answer, explaining the weird error.
The model limit is 10gb.
Hii @enzo, can you please tell us how you are calculating this spearman correlation. is spearman calculated for each gene or it is normally used by scipy.
The score is a weighted average of a cell-wise Spearman.
The weights come from a file provided by broad, which regroups the cells before doing the mean.
The final score is an average of all tissue samples.