{ "id": 264799, "title": "Submission error", "url": "/competitions/landmark-recognition-2021/discussion/264799", "author_name": "DrCapa", "post_date": "2021-08-13T09:32:08.139000", "votes": 13, "comment_count": 16, "views": 0, "content": "

Hi @all,

\n

unfortunately i get an submission error without informations with respect to the reason. My notebook is here.

\n

An this is the error message: \"\"

\n

Has somebody a helpful advice?

\n

Best regards
\nCapa

", "messages": [ { "id": 1470092, "postDate": "2021-08-13T09:32:08.140Z", "content": "

Hi @all,

\n

unfortunately i get an submission error without informations with respect to the reason. My notebook is here.

\n

An this is the error message: \"\"

\n

Has somebody a helpful advice?

\n

Best regards
\nCapa

", "rawMarkdown": "Hi @all,\n\nunfortunately i get an submission error without informations with respect to the reason. My notebook is [here](https://www.kaggle.com/drcapa/landmark-recognition-2021-starter).\n\nAn this is the error message: ![](https://i.ibb.co/Lv411qp/sub-error.png )\n\nHas somebody a helpful advice?\n\nBest regards\nCapa", "votes": 13 }, { "id": 1485973, "postDate": "2021-08-22T14:48:16.807Z", "content": "

Hey, @drcapa! I faced the same issue today. The reason was that I was trying to read train.csv on my inference (in my case for one-hot decoding). Just remove any requests to the train.csv or train folder, and you'll fix the issue.

", "rawMarkdown": "Hey, @drcapa! I faced the same issue today. The reason was that I was trying to read `train.csv` on my inference (in my case for one-hot decoding). Just remove any requests to the `train.csv` or `train` folder, and you'll fix the issue.", "votes": 4 }, { "id": 1474073, "postDate": "2021-08-15T22:28:29.423Z", "content": "

Check that you do not use in the script any of the images that in the private rerun might not be among the 100k selected by the host.

", "rawMarkdown": "Check that you do not use in the script any of the images that in the private rerun might not be among the 100k selected by the host.", "votes": 4 }, { "id": 1505065, "postDate": "2021-09-06T22:19:43.973Z", "content": "

I'm experiencing the same issues 🤕. It perfectly runs on public test set, but raises an error on the private one.

\n

I'm doing the following blind testing:

\n\n

Repeat it until it stop launching errors, to detect the first line causing an error.

\n

You can remove the very last line, or perform a dichotomic search 😩

", "rawMarkdown": "I'm experiencing the same issues 🤕. It perfectly runs on public test set, but raises an error on the private one.\n\nI'm doing the following blind testing:\n\n- start the script with: \n `!cp ../input/landmark-recognition-2021/sample_submission.csv submission.csv`\n\n- then remove the very last lines & execute\n\nRepeat it until it stop launching errors, to detect the first line causing an error.\n\nYou can remove the very last line, or perform a dichotomic search 😩", "votes": 1 }, { "id": 1478738, "postDate": "2021-08-18T05:57:18.943Z", "content": "

Hey, I have faces similar issue while submitting. You can try few things:

\n
    \n
  1. Check the index/Name closely of submission.csv.
    2. \n
  2. Did you train model inside notebooks? Try to generate submission.csv inside kaggle notebooks
    3. \n
  3. Save the notebooks. Sometime due to bad internet connections, expected notebook did not get saved properly.
    4. \n
", "rawMarkdown": "Hey, I have faces similar issue while submitting. You can try few things:\n1. Check the index/Name closely of submission.csv. \n2. Did you train model inside notebooks? Try to generate submission.csv inside kaggle notebooks\n3. Save the notebooks. Sometime due to bad internet connections, expected notebook did not get saved properly.", "votes": 1 }, { "id": 1475228, "postDate": "2021-08-16T14:17:03.003Z", "content": "

To those seeking to debug submission errors, please view https://www.kaggle.com/code-competition-debugging

", "rawMarkdown": "To those seeking to debug submission errors, please view https://www.kaggle.com/code-competition-debugging ", "votes": 1, "replies": [ { "id": 1487275, "postDate": "2021-08-23T14:11:56.443Z", "content": "

Hi @wcukierski in the debugging tips page, it says \"Notebook Threw Exception\" only happens when running scripts but not notebooks, but I keep getting this error when submitting notebooks. Look like it's the same case for OP. Any idea why?

\n
\n

Notebook Threw Exception: While rerunning your code, your notebook hit an unhandled error. Note that the hidden dataset can be larger/smaller/different than the public dataset. Due to the way Kaggle handles notebook runtime exceptions, this error class is currently only possible with scripts (not notebooks).

\n
", "rawMarkdown": "Hi @wcukierski in the debugging tips page, it says \"Notebook Threw Exception\" only happens when running scripts but not notebooks, but I keep getting this error when submitting notebooks. Look like it's the same case for OP. Any idea why?\n\n> Notebook Threw Exception: While rerunning your code, your notebook hit an unhandled error. Note that the hidden dataset can be larger/smaller/different than the public dataset. Due to the way Kaggle handles notebook runtime exceptions, this error class is currently only possible with scripts (not notebooks).\n", "votes": 1 }, { "id": 1487367, "postDate": "2021-08-23T15:24:14.750Z", "content": "

@boliu0 Thanks for pointing that out. That text is out of date and no longer true. I'll update the page.

", "rawMarkdown": "@boliu0 Thanks for pointing that out. That text is out of date and no longer true. I'll update the page.", "votes": 2 } ] }, { "id": 1470229, "postDate": "2021-08-13T11:42:40.487Z", "content": "

Get the same error, didnt know why, because predictions looks correct

", "rawMarkdown": "Get the same error, didnt know why, because predictions looks correct", "votes": 2, "replies": [ { "id": 1490629, "postDate": "2021-08-25T18:24:35.150Z", "content": "

Seems like you have fixed the problem, how have you dealt with this issue?

", "rawMarkdown": "Seems like you have fixed the problem, how have you dealt with this issue?" }, { "id": 1491039, "postDate": "2021-08-26T05:30:11.827Z", "content": "

No, I didnt, I submited other notebook

", "rawMarkdown": "No, I didnt, I submited other notebook" } ] }, { "id": 1471476, "postDate": "2021-08-14T08:27:03.767Z", "content": "

I got the same \"Submission Scoring Error\" the first time because my Internet was on and when I tried by uploading the extra libraries via upload it threw a \"Notebook Threw Exception\" during submission evaluation, The notebooks seem to be working fine when I use the \"save version\" function, not sure exactly why this is happening.

", "rawMarkdown": "I got the same \"Submission Scoring Error\" the first time because my Internet was on and when I tried by uploading the extra libraries via upload it threw a \"Notebook Threw Exception\" during submission evaluation, The notebooks seem to be working fine when I use the \"save version\" function, not sure exactly why this is happening." }, { "id": 1506218, "postDate": "2021-09-08T03:09:27.150Z", "content": "

Maybe it's tensorflow's problem. I have observed memory leaks when inferring, reducing the batch size or inferring image by image may alleviate the situation. I haven't found a perfect solution yet.

", "rawMarkdown": "Maybe it's tensorflow's problem. I have observed memory leaks when inferring, reducing the batch size or inferring image by image may alleviate the situation. I haven't found a perfect solution yet." }, { "id": 1486630, "postDate": "2021-08-23T05:41:23.890Z", "content": "

Do we need to send the notebook in which we have trained the model?
\nOr can I load the model that I've just trained?
\nMoreover, is it acceptable to save numpy array as .h5 format and load it to my evaluation notebook?

", "rawMarkdown": "Do we need to send the notebook in which we have trained the model?\nOr can I load the model that I've just trained?\nMoreover, is it acceptable to save numpy array as .h5 format and load it to my evaluation notebook?" }, { "id": 1481018, "postDate": "2021-08-19T09:17:03.887Z", "content": "

Get the same error

", "rawMarkdown": "Get the same error", "replies": [ { "id": 1490627, "postDate": "2021-08-25T18:23:57.220Z", "content": "

How you've managed to deal with this problem? Do you use the train folder in the hidden Dataset?

", "rawMarkdown": "How you've managed to deal with this problem? Do you use the train folder in the hidden Dataset?" } ] }, { "id": 1470918, "postDate": "2021-08-13T19:37:45.647Z", "content": "

This usually happens when there is a difference between the public and hidden test set, that is your train generator might be broken. I would recommend doing the following:

\n
    \n
  1. Build your model without training -> make a submission where only noise is passed through as predictions.
    2. \n
  2. Build your model with training -> make a submission where noise is passed through as predictions.
    3. \n
  3. Build your model without training -> make a submission where the actual test data is passed through as predictions.
    4. \n
  4. Build your model with training -> make a submission where the actual test data is passed through as predictions.
    5. \n
\n

This should give you a good idea of what part of your notebook is erroring. You can probably get away with just loading weights as training to save some time.

", "rawMarkdown": "This usually happens when there is a difference between the public and hidden test set, that is your train generator might be broken. I would recommend doing the following:\n\n1. Build your model without training -> make a submission where only noise is passed through as predictions.\n2. Build your model with training -> make a submission where noise is passed through as predictions.\n3. Build your model without training -> make a submission where the actual test data is passed through as predictions.\n4. Build your model with training -> make a submission where the actual test data is passed through as predictions.\n\nThis should give you a good idea of what part of your notebook is erroring. You can probably get away with just loading weights as training to save some time.", "votes": 2, "isDeleted": true } ], "comments": [ { "id": 1485973, "author_name": "Vadim Timakin", "author_url": "", "post_date": "2021-08-22T14:48:16.807000", "content": "

Hey, @drcapa! I faced the same issue today. The reason was that I was trying to read train.csv on my inference (in my case for one-hot decoding). Just remove any requests to the train.csv or train folder, and you'll fix the issue.

", "votes": 4, "replies": [] }, { "id": 1474073, "author_name": "Manuel Campos", "author_url": "", "post_date": "2021-08-15T22:28:29.423000", "content": "

Check that you do not use in the script any of the images that in the private rerun might not be among the 100k selected by the host.

", "votes": 4, "replies": [] }, { "id": 1505065, "author_name": "Virilo Tejedor Aguilera", "author_url": "", "post_date": "2021-09-06T22:19:43.973000", "content": "

I'm experiencing the same issues 🤕. It perfectly runs on public test set, but raises an error on the private one.

\n

I'm doing the following blind testing:

\n\n

Repeat it until it stop launching errors, to detect the first line causing an error.

\n

You can remove the very last line, or perform a dichotomic search 😩

", "votes": 1, "replies": [] }, { "id": 1478738, "author_name": "Saurav Solanki", "author_url": "", "post_date": "2021-08-18T05:57:18.943000", "content": "

Hey, I have faces similar issue while submitting. You can try few things:

\n
    \n
  1. Check the index/Name closely of submission.csv.
    2. \n
  2. Did you train model inside notebooks? Try to generate submission.csv inside kaggle notebooks
    3. \n
  3. Save the notebooks. Sometime due to bad internet connections, expected notebook did not get saved properly.
    4. \n
", "votes": 1, "replies": [] }, { "id": 1475228, "author_name": "Will Cukierski", "author_url": "", "post_date": "2021-08-16T14:17:03.003000", "content": "

To those seeking to debug submission errors, please view https://www.kaggle.com/code-competition-debugging

", "votes": 1, "replies": [ { "id": 1487275, "author_name": "Bo", "author_url": "", "post_date": "2021-08-23T14:11:56.443000", "content": "

Hi @wcukierski in the debugging tips page, it says \"Notebook Threw Exception\" only happens when running scripts but not notebooks, but I keep getting this error when submitting notebooks. Look like it's the same case for OP. Any idea why?

\n
\n

Notebook Threw Exception: While rerunning your code, your notebook hit an unhandled error. Note that the hidden dataset can be larger/smaller/different than the public dataset. Due to the way Kaggle handles notebook runtime exceptions, this error class is currently only possible with scripts (not notebooks).

\n
", "votes": 1, "replies": [] }, { "id": 1487367, "author_name": "Will Cukierski", "author_url": "", "post_date": "2021-08-23T15:24:14.750000", "content": "

@boliu0 Thanks for pointing that out. That text is out of date and no longer true. I'll update the page.

", "votes": 2, "replies": [] } ] }, { "id": 1470229, "author_name": "Radmir Zosimov", "author_url": "", "post_date": "2021-08-13T11:42:40.487000", "content": "

Get the same error, didnt know why, because predictions looks correct

", "votes": 2, "replies": [ { "id": 1490629, "author_name": "nirzaa", "author_url": "", "post_date": "2021-08-25T18:24:35.150000", "content": "

Seems like you have fixed the problem, how have you dealt with this issue?

", "votes": 0, "replies": [] }, { "id": 1491039, "author_name": "Radmir Zosimov", "author_url": "", "post_date": "2021-08-26T05:30:11.827000", "content": "

No, I didnt, I submited other notebook

", "votes": 0, "replies": [] } ] }, { "id": 1471476, "author_name": "Pranav Singh", "author_url": "", "post_date": "2021-08-14T08:27:03.767000", "content": "

I got the same \"Submission Scoring Error\" the first time because my Internet was on and when I tried by uploading the extra libraries via upload it threw a \"Notebook Threw Exception\" during submission evaluation, The notebooks seem to be working fine when I use the \"save version\" function, not sure exactly why this is happening.

", "votes": 0, "replies": [] }, { "id": 1506218, "author_name": "Qiao Su", "author_url": "", "post_date": "2021-09-08T03:09:27.150000", "content": "

Maybe it's tensorflow's problem. I have observed memory leaks when inferring, reducing the batch size or inferring image by image may alleviate the situation. I haven't found a perfect solution yet.

", "votes": 0, "replies": [] }, { "id": 1486630, "author_name": "nirzaa", "author_url": "", "post_date": "2021-08-23T05:41:23.890000", "content": "

Do we need to send the notebook in which we have trained the model?
\nOr can I load the model that I've just trained?
\nMoreover, is it acceptable to save numpy array as .h5 format and load it to my evaluation notebook?

", "votes": 0, "replies": [] }, { "id": 1481018, "author_name": "Chrisfsj2051", "author_url": "", "post_date": "2021-08-19T09:17:03.887000", "content": "

Get the same error

", "votes": 0, "replies": [ { "id": 1490627, "author_name": "nirzaa", "author_url": "", "post_date": "2021-08-25T18:23:57.220000", "content": "

How you've managed to deal with this problem? Do you use the train folder in the hidden Dataset?

", "votes": 0, "replies": [] } ] }, { "id": 1470918, "author_name": "", "author_url": "", "post_date": "2021-08-13T19:37:45.647000", "content": "

This usually happens when there is a difference between the public and hidden test set, that is your train generator might be broken. I would recommend doing the following:

\n
    \n
  1. Build your model without training -> make a submission where only noise is passed through as predictions.
    2. \n
  2. Build your model with training -> make a submission where noise is passed through as predictions.
    3. \n
  3. Build your model without training -> make a submission where the actual test data is passed through as predictions.
    4. \n
  4. Build your model with training -> make a submission where the actual test data is passed through as predictions.
    5. \n
\n

This should give you a good idea of what part of your notebook is erroring. You can probably get away with just loading weights as training to save some time.

", "votes": 2, "replies": [] } ], "raw_markdown_by_id": { "1470092": "Hi @all,\n\nunfortunately i get an submission error without informations with respect to the reason. My notebook is [here](https://www.kaggle.com/drcapa/landmark-recognition-2021-starter).\n\nAn this is the error message: ![](https://i.ibb.co/Lv411qp/sub-error.png )\n\nHas somebody a helpful advice?\n\nBest regards\nCapa", "1485973": "Hey, @drcapa! I faced the same issue today. The reason was that I was trying to read `train.csv` on my inference (in my case for one-hot decoding). Just remove any requests to the `train.csv` or `train` folder, and you'll fix the issue.", "1474073": "Check that you do not use in the script any of the images that in the private rerun might not be among the 100k selected by the host.", "1505065": "I'm experiencing the same issues 🤕. It perfectly runs on public test set, but raises an error on the private one.\n\nI'm doing the following blind testing:\n\n- start the script with: \n `!cp ../input/landmark-recognition-2021/sample_submission.csv submission.csv`\n\n- then remove the very last lines & execute\n\nRepeat it until it stop launching errors, to detect the first line causing an error.\n\nYou can remove the very last line, or perform a dichotomic search 😩", "1478738": "Hey, I have faces similar issue while submitting. You can try few things:\n1. Check the index/Name closely of submission.csv. \n2. Did you train model inside notebooks? Try to generate submission.csv inside kaggle notebooks\n3. Save the notebooks. Sometime due to bad internet connections, expected notebook did not get saved properly.", "1475228": "To those seeking to debug submission errors, please view https://www.kaggle.com/code-competition-debugging ", "1470229": "Get the same error, didnt know why, because predictions looks correct", "1471476": "I got the same \"Submission Scoring Error\" the first time because my Internet was on and when I tried by uploading the extra libraries via upload it threw a \"Notebook Threw Exception\" during submission evaluation, The notebooks seem to be working fine when I use the \"save version\" function, not sure exactly why this is happening.", "1506218": "Maybe it's tensorflow's problem. I have observed memory leaks when inferring, reducing the batch size or inferring image by image may alleviate the situation. I haven't found a perfect solution yet.", "1486630": "Do we need to send the notebook in which we have trained the model?\nOr can I load the model that I've just trained?\nMoreover, is it acceptable to save numpy array as .h5 format and load it to my evaluation notebook?", "1481018": "Get the same error", "1470918": "This usually happens when there is a difference between the public and hidden test set, that is your train generator might be broken. I would recommend doing the following:\n\n1. Build your model without training -> make a submission where only noise is passed through as predictions.\n2. Build your model with training -> make a submission where noise is passed through as predictions.\n3. Build your model without training -> make a submission where the actual test data is passed through as predictions.\n4. Build your model with training -> make a submission where the actual test data is passed through as predictions.\n\nThis should give you a good idea of what part of your notebook is erroring. You can probably get away with just loading weights as training to save some time." } }