Currently getting a submission scoring error on my submissions and I can't quite figure out why. Tried some of the things I saw in past discussion threads, but still can't get it to work. I'd really appreciate if someone could troubleshoot and tell me what's wrong!
\nNotebook link: https://www.kaggle.com/bharatraghavan03/rsna2024-inference-nb
", "messages": [ { "id": "2989996", "postDate": "09/15/2024 19:11:47", "content": "Currently getting a submission scoring error on my submissions and I can't quite figure out why. Tried some of the things I saw in past discussion threads, but still can't get it to work. I'd really appreciate if someone could troubleshoot and tell me what's wrong!
\nNotebook link: https://www.kaggle.com/bharatraghavan03/rsna2024-inference-nb
", "rawMarkdown": "Currently getting a submission scoring error on my submissions and I can't quite figure out why. Tried some of the things I saw in past discussion threads, but still can't get it to work. I'd really appreciate if someone could troubleshoot and tell me what's wrong!\n\nNotebook link: [https://www.kaggle.com/bharatraghavan03/rsna2024-inference-nb](https://www.kaggle.com/bharatraghavan03/rsna2024-inference-nb)", "votes": null }, { "id": "2990059", "postDate": "09/15/2024 21:38:22", "content": "If I understand correctly you are adding row_names for each series_id. Some study_id can have multiple series_id for the same type of MRI or even none. The local test works because has one of each.
\nA submission_df = submission_df.groupby('row_id').mean() before sample_submission.merge(...) should work for multiples.
Omg that's a great point, will try changing it (after my submissions reset lol) and I'll let you know. Thanks so much regardless!
", "rawMarkdown": "Omg that's a great point, will try changing it (after my submissions reset lol) and I'll let you know. Thanks so much regardless!", "votes": null }, { "id": "2990062", "postDate": "09/15/2024 21:51:16", "content": "I've just saw .fillna(-100)
I think you should use \"1./3\" instead and may be recheck all predictions sum 1.
", "rawMarkdown": "I've just saw `.fillna(-100)`\n\nI think you should use \"1./3\" instead and may be recheck all predictions sum 1.", "votes": null }, { "id": "2990107", "postDate": "09/16/2024 01:11:02", "content": "I think the groupby (or lack thereof) was the problem; just submitted and it worked. Thank you so much!
i was wondering ,if it happens that there is a series with a missing condition or level ,will your code be still an accurate solution to describe row_ids ?.Or the row id is not that significant to the scoring or there are no missing conditions in the test set ? can you explain
", "rawMarkdown": "i was wondering ,if it happens that there is a series with a missing condition or level ,will your code be still an accurate solution to describe row_ids ?.Or the row id is not that significant to the scoring or there are no missing conditions in the test set ? can you explain", "votes": null }, { "id": "2991023", "postDate": "09/17/2024 00:02:19", "content": "I think that might've been a problem if I didn't merge at the end. Because it's a left merge, we preserve every row required in the submission, and just fill anything that isn't merged (i.e. rows that are not present in the predictions).
", "rawMarkdown": "I think that might've been a problem if I didn't merge at the end. Because it's a left merge, we preserve every row required in the submission, and just fill anything that isn't merged (i.e. rows that are not present in the predictions).", "votes": null }, { "id": "2992897", "postDate": "09/19/2024 08:21:56", "content": "take a look at discussion here
", "rawMarkdown": "take a look at discussion [here](https://www.kaggle.com/competitions/rsna-2024-lumbar-spine-degenerative-classification/discussion/523247)", "votes": null } ], "comments": [ { "id": 2990059, "author_name": "sacuscreed", "author_url": "", "post_date": "09/15/2024 21:38:22", "content": "If I understand correctly you are adding row_names for each series_id. Some study_id can have multiple series_id for the same type of MRI or even none. The local test works because has one of each.
\nA submission_df = submission_df.groupby('row_id').mean() before sample_submission.merge(...) should work for multiples.
Omg that's a great point, will try changing it (after my submissions reset lol) and I'll let you know. Thanks so much regardless!
", "votes": null, "replies": [ { "id": 2990062, "author_name": "sacuscreed", "author_url": "", "post_date": "09/15/2024 21:51:16", "content": "I've just saw .fillna(-100)
I think you should use \"1./3\" instead and may be recheck all predictions sum 1.
", "votes": null, "replies": [ { "id": 2990107, "author_name": "bharatraghavan03", "author_url": "", "post_date": "09/16/2024 01:11:02", "content": "I think the groupby (or lack thereof) was the problem; just submitted and it worked. Thank you so much!
i was wondering ,if it happens that there is a series with a missing condition or level ,will your code be still an accurate solution to describe row_ids ?.Or the row id is not that significant to the scoring or there are no missing conditions in the test set ? can you explain
", "votes": null, "replies": [ { "id": 2991023, "author_name": "bharatraghavan03", "author_url": "", "post_date": "09/17/2024 00:02:19", "content": "I think that might've been a problem if I didn't merge at the end. Because it's a left merge, we preserve every row required in the submission, and just fill anything that isn't merged (i.e. rows that are not present in the predictions).
", "votes": null, "replies": [] } ] } ] } ] } ] } ] }, { "id": 2992897, "author_name": "rohitchaudhari25", "author_url": "", "post_date": "09/19/2024 08:21:56", "content": "take a look at discussion here
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2989996": "Currently getting a submission scoring error on my submissions and I can't quite figure out why. Tried some of the things I saw in past discussion threads, but still can't get it to work. I'd really appreciate if someone could troubleshoot and tell me what's wrong!\n\nNotebook link: [https://www.kaggle.com/bharatraghavan03/rsna2024-inference-nb](https://www.kaggle.com/bharatraghavan03/rsna2024-inference-nb)", "2990059": "If I understand correctly you are adding row_names for each series_id. Some study_id can have multiple series_id for the same type of MRI or even none. The local test works because has one of each.\n\nA `submission_df = submission_df.groupby('row_id').mean()` before `sample_submission.merge(...)` should work for multiples.", "2990060": "Omg that's a great point, will try changing it (after my submissions reset lol) and I'll let you know. Thanks so much regardless!", "2990062": "I've just saw `.fillna(-100)`\n\nI think you should use \"1./3\" instead and may be recheck all predictions sum 1.", "2990107": "I think the `groupby` (or lack thereof) was the problem; just submitted and it worked. Thank you so much!", "2990684": "i was wondering ,if it happens that there is a series with a missing condition or level ,will your code be still an accurate solution to describe row_ids ?.Or the row id is not that significant to the scoring or there are no missing conditions in the test set ? can you explain", "2991023": "I think that might've been a problem if I didn't merge at the end. Because it's a left merge, we preserve every row required in the submission, and just fill anything that isn't merged (i.e. rows that are not present in the predictions).", "2992897": "take a look at discussion [here](https://www.kaggle.com/competitions/rsna-2024-lumbar-spine-degenerative-classification/discussion/523247)" }, "source": "meta" }