{ "id": 514655, "title": "Scoring error", "url": "/competitions/rsna-2024-lumbar-spine-degenerative-classification/discussion/514655", "author_name": "", "post_date": "2024-06-25T04:03:48.724435800Z", "votes": 1, "comment_count": 11, "views": 0, "content": "

I can't manage to get my submission score. I know it has the right columns, right number of rows and, the columns sum to 1, all the 'row_id' values are good.
\nI need to know if the error could be coming from the fact that the rows must be in a particular order. Also, if it's not the case, what could cause a scoring error ?
\nI leave the head of the submission dataframe (training data).
\n

", "messages": [ { "id": "2888751", "postDate": "06/25/2024 04:03:48", "content": "

I can't manage to get my submission score. I know it has the right columns, right number of rows and, the columns sum to 1, all the 'row_id' values are good.
\nI need to know if the error could be coming from the fact that the rows must be in a particular order. Also, if it's not the case, what could cause a scoring error ?
\nI leave the head of the submission dataframe (training data).
\n

", "rawMarkdown": "I can't manage to get my submission score. I know it has the right columns, right number of rows and, the columns sum to 1, all the 'row_id' values are good. \nI need to know if the error could be coming from the fact that the rows must be in a particular order. Also, if it's not the case, what could cause a scoring error ? \nI leave the head of the submission dataframe (training data). \n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F14858850%2F6abdea3a5b804b6772c0145d7a57995b%2FScreenshot%202024-06-24%20at%2020.55.43.png?generation=1719288201549913&alt=media)", "votes": null }, { "id": "2889431", "postDate": "06/25/2024 13:42:49", "content": "

Are you saving the DataFrame with or without index? The correct way is like this: my_df.to_csv(\"/kaggle/working/submission.csv\", index=False)

", "rawMarkdown": "Are you saving the DataFrame with or without index? The correct way is like this: `my_df.to_csv(\"/kaggle/working/submission.csv\", index=False)`", "votes": null }, { "id": "2889689", "postDate": "06/25/2024 16:14:51", "content": "

Thank you for your response. Yes, this is the way I save the dataframe.

", "rawMarkdown": "Thank you for your response. Yes, this is the way I save the dataframe.", "votes": null }, { "id": "2890284", "postDate": "06/26/2024 04:02:07", "content": "

Here are some things you can check:

\n
    \n
  1. Assert that all row_ids in sample_submission.csv are present in your final dataframe and that there are no additional row_ids. That is, set(ss_df.row_id.to_list())==set(my_df.row_id.to_list()) where my_df is your final submission dataframe.
    2. \n
  2. Make sure row_ids are unique.
    3. \n
  3. Round the floats, to say 6 decimals, as exponential notations can cause error.
    4. \n
  4. I don't think row_ids should be ordered in the same as sample submission, but if that is a problem, use this code: my_df= ss_df.merge(my_df, on=\"row_id\", validate=\"1:1:). Afterwards assert that there are no nan values.
    5. \n
\n

If the \"Submission Scoring Error\" changes to \"Notebook Throw an Exception\" then you know that some problem exists in your code.

\n

Also see my starter notebook where I used a flag to replace the test data with a sample selected from the train set. The reason for the flag (len(sub) <= 25) so that when submitting it for scoring, it will be false and the actual test data will be used.

", "rawMarkdown": "Here are some things you can check:\n1. Assert that all `row_ids` in `sample_submission.csv` are present in your final dataframe and that there are no additional `row_ids`. That is, `set(ss_df.row_id.to_list())==set(my_df.row_id.to_list())` where `my_df` is your final submission dataframe.\n1. Make sure `row_ids` are unique.\n1. Round the floats, to say 6 decimals, as exponential notations can cause error.\n1. I don't think `row_ids` should be ordered in the same as sample submission, but if that is a problem, use this code: `my_df= ss_df.merge(my_df, on=\"row_id\", validate=\"1:1:)`. Afterwards assert that there are no nan values.\n\nIf the \"Submission Scoring Error\" changes to \"Notebook Throw an Exception\" then you know that some problem exists in your code.\n\nAlso see my [starter notebook](https://www.kaggle.com/code/coderrkj/rsna-resnet-starter-notebook) where I used a flag to replace the test data with a sample selected from the train set. The reason for the flag (`len(sub) <= 25`) so that when submitting it for scoring, it will be false and the actual test data will be used.", "votes": null }, { "id": "2890625", "postDate": "06/26/2024 07:40:09", "content": "

Have you solved the problem? I had the same problem this morning

", "rawMarkdown": "Have you solved the problem? I had the same problem this morning", "votes": null }, { "id": "2900398", "postDate": "07/02/2024 08:51:16", "content": "

Could you give me some advice for the related topic (https://www.kaggle.com/competitions/rsna-2024-lumbar-spine-degenerative-classification/discussion/516357). Thanks!

", "rawMarkdown": "Could you give me some advice for the related topic (https://www.kaggle.com/competitions/rsna-2024-lumbar-spine-degenerative-classification/discussion/516357). Thanks!", "votes": null }, { "id": "2900399", "postDate": "07/02/2024 08:52:01", "content": "

Have you tackled this problem?

", "rawMarkdown": "Have you tackled this problem?", "votes": null }, { "id": "2916921", "postDate": "07/11/2024 11:03:33", "content": "

Thanks for your suggestion, however, it would interest you to know that the sample_submission is not part of the training datasets given. Kindly check

", "rawMarkdown": "Thanks for your suggestion, however, it would interest you to know that the sample_submission is not part of the training datasets given. Kindly check", "votes": null }, { "id": "2938923", "postDate": "07/28/2024 15:35:42", "content": "

I am facing same issue, did you got any leads ?
\nis it necessary that all rows sum up to 1?
\nthere may be rounding errors etc.

", "rawMarkdown": "I am facing same issue, did you got any leads ?\nis it necessary that all rows sum up to 1?\nthere may be rounding errors etc.", "votes": null }, { "id": "2938991", "postDate": "07/28/2024 16:38:19", "content": "

Summing up to 1 is not strictly necessary but all values need to be non-negative.
\nTrying to check all the 4 points I have listed below in one of the comments.

", "rawMarkdown": "Summing up to 1 is not strictly necessary but all values need to be non-negative.\nTrying to check all the 4 points I have listed below in one of the comments.", "votes": null }, { "id": "2939936", "postDate": "07/29/2024 16:04:15", "content": "

I have a doubt.
\nDoes kaggle update sample_submission.csv file as well for code competitions with hidden test set at the time of scoring?

\n

If not then asserting that the row_ids set in sample_submission.csv and submission.csv are same does not make sense.

", "rawMarkdown": "I have a doubt.\nDoes kaggle update sample_submission.csv file as well for code competitions with hidden test set at the time of scoring?\n\nIf not then asserting that the row_ids set in sample_submission.csv and submission.csv are same does not make sense.", "votes": null }, { "id": "2940038", "postDate": "07/29/2024 17:57:29", "content": "

It does, at least for this competition. See one of the baseline notebooks initially published at the start of this completion: https://www.kaggle.com/code/krabhijeet/rsna2024-baseline-submission

\n

They are using sample_submission.csv and nothing else to get a public score around 1.

", "rawMarkdown": "It does, at least for this competition. See one of the baseline notebooks initially published at the start of this completion: https://www.kaggle.com/code/krabhijeet/rsna2024-baseline-submission\n\nThey are using `sample_submission.csv` and nothing else to get a public score around 1.", "votes": null } ], "comments": [ { "id": 2889431, "author_name": "coderrkj", "author_url": "", "post_date": "06/25/2024 13:42:49", "content": "

Are you saving the DataFrame with or without index? The correct way is like this: my_df.to_csv(\"/kaggle/working/submission.csv\", index=False)

", "votes": null, "replies": [ { "id": 2889689, "author_name": "juliengenzling", "author_url": "", "post_date": "06/25/2024 16:14:51", "content": "

Thank you for your response. Yes, this is the way I save the dataframe.

", "votes": null, "replies": [] } ] }, { "id": 2890284, "author_name": "coderrkj", "author_url": "", "post_date": "06/26/2024 04:02:07", "content": "

Here are some things you can check:

\n
    \n
  1. Assert that all row_ids in sample_submission.csv are present in your final dataframe and that there are no additional row_ids. That is, set(ss_df.row_id.to_list())==set(my_df.row_id.to_list()) where my_df is your final submission dataframe.
    2. \n
  2. Make sure row_ids are unique.
    3. \n
  3. Round the floats, to say 6 decimals, as exponential notations can cause error.
    4. \n
  4. I don't think row_ids should be ordered in the same as sample submission, but if that is a problem, use this code: my_df= ss_df.merge(my_df, on=\"row_id\", validate=\"1:1:). Afterwards assert that there are no nan values.
    5. \n
\n

If the \"Submission Scoring Error\" changes to \"Notebook Throw an Exception\" then you know that some problem exists in your code.

\n

Also see my starter notebook where I used a flag to replace the test data with a sample selected from the train set. The reason for the flag (len(sub) <= 25) so that when submitting it for scoring, it will be false and the actual test data will be used.

", "votes": null, "replies": [ { "id": 2900398, "author_name": "tuanle98", "author_url": "", "post_date": "07/02/2024 08:51:16", "content": "

Could you give me some advice for the related topic (https://www.kaggle.com/competitions/rsna-2024-lumbar-spine-degenerative-classification/discussion/516357). Thanks!

", "votes": null, "replies": [] }, { "id": 2916921, "author_name": "samsonolutoberu", "author_url": "", "post_date": "07/11/2024 11:03:33", "content": "

Thanks for your suggestion, however, it would interest you to know that the sample_submission is not part of the training datasets given. Kindly check

", "votes": null, "replies": [] } ] }, { "id": 2890625, "author_name": "", "author_url": "", "post_date": "06/26/2024 07:40:09", "content": "

Have you solved the problem? I had the same problem this morning

", "votes": null, "replies": [] }, { "id": 2900399, "author_name": "tuanle98", "author_url": "", "post_date": "07/02/2024 08:52:01", "content": "

Have you tackled this problem?

", "votes": null, "replies": [] }, { "id": 2938923, "author_name": "rohitchaudhari25", "author_url": "", "post_date": "07/28/2024 15:35:42", "content": "

I am facing same issue, did you got any leads ?
\nis it necessary that all rows sum up to 1?
\nthere may be rounding errors etc.

", "votes": null, "replies": [ { "id": 2938991, "author_name": "coderrkj", "author_url": "", "post_date": "07/28/2024 16:38:19", "content": "

Summing up to 1 is not strictly necessary but all values need to be non-negative.
\nTrying to check all the 4 points I have listed below in one of the comments.

", "votes": null, "replies": [ { "id": 2939936, "author_name": "rohitchaudhari25", "author_url": "", "post_date": "07/29/2024 16:04:15", "content": "

I have a doubt.
\nDoes kaggle update sample_submission.csv file as well for code competitions with hidden test set at the time of scoring?

\n

If not then asserting that the row_ids set in sample_submission.csv and submission.csv are same does not make sense.

", "votes": null, "replies": [ { "id": 2940038, "author_name": "coderrkj", "author_url": "", "post_date": "07/29/2024 17:57:29", "content": "

It does, at least for this competition. See one of the baseline notebooks initially published at the start of this completion: https://www.kaggle.com/code/krabhijeet/rsna2024-baseline-submission

\n

They are using sample_submission.csv and nothing else to get a public score around 1.

", "votes": null, "replies": [] } ] } ] } ] } ], "raw_markdown_by_id": { "2888751": "I can't manage to get my submission score. I know it has the right columns, right number of rows and, the columns sum to 1, all the 'row_id' values are good. \nI need to know if the error could be coming from the fact that the rows must be in a particular order. Also, if it's not the case, what could cause a scoring error ? \nI leave the head of the submission dataframe (training data). \n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F14858850%2F6abdea3a5b804b6772c0145d7a57995b%2FScreenshot%202024-06-24%20at%2020.55.43.png?generation=1719288201549913&alt=media)", "2889431": "Are you saving the DataFrame with or without index? The correct way is like this: `my_df.to_csv(\"/kaggle/working/submission.csv\", index=False)`", "2889689": "Thank you for your response. Yes, this is the way I save the dataframe.", "2890284": "Here are some things you can check:\n1. Assert that all `row_ids` in `sample_submission.csv` are present in your final dataframe and that there are no additional `row_ids`. That is, `set(ss_df.row_id.to_list())==set(my_df.row_id.to_list())` where `my_df` is your final submission dataframe.\n1. Make sure `row_ids` are unique.\n1. Round the floats, to say 6 decimals, as exponential notations can cause error.\n1. I don't think `row_ids` should be ordered in the same as sample submission, but if that is a problem, use this code: `my_df= ss_df.merge(my_df, on=\"row_id\", validate=\"1:1:)`. Afterwards assert that there are no nan values.\n\nIf the \"Submission Scoring Error\" changes to \"Notebook Throw an Exception\" then you know that some problem exists in your code.\n\nAlso see my [starter notebook](https://www.kaggle.com/code/coderrkj/rsna-resnet-starter-notebook) where I used a flag to replace the test data with a sample selected from the train set. The reason for the flag (`len(sub) <= 25`) so that when submitting it for scoring, it will be false and the actual test data will be used.", "2890625": "Have you solved the problem? I had the same problem this morning", "2900398": "Could you give me some advice for the related topic (https://www.kaggle.com/competitions/rsna-2024-lumbar-spine-degenerative-classification/discussion/516357). Thanks!", "2900399": "Have you tackled this problem?", "2916921": "Thanks for your suggestion, however, it would interest you to know that the sample_submission is not part of the training datasets given. Kindly check", "2938923": "I am facing same issue, did you got any leads ?\nis it necessary that all rows sum up to 1?\nthere may be rounding errors etc.", "2938991": "Summing up to 1 is not strictly necessary but all values need to be non-negative.\nTrying to check all the 4 points I have listed below in one of the comments.", "2939936": "I have a doubt.\nDoes kaggle update sample_submission.csv file as well for code competitions with hidden test set at the time of scoring?\n\nIf not then asserting that the row_ids set in sample_submission.csv and submission.csv are same does not make sense.", "2940038": "It does, at least for this competition. See one of the baseline notebooks initially published at the start of this completion: https://www.kaggle.com/code/krabhijeet/rsna2024-baseline-submission\n\nThey are using `sample_submission.csv` and nothing else to get a public score around 1." }, "source": "meta" }