{ "id": 569027, "title": "Scoring Error on ECAPA TDNN", "url": "/competitions/birdclef-2025/discussion/569027", "author_name": "", "post_date": "2025-03-19T13:30:29.928295800Z", "votes": 1, "comment_count": 14, "views": 0, "content": "

Hi everybody
\nthis is my solution Notebook
\nI have no idea why I got submission scoring error

", "messages": [ { "id": "3154069", "postDate": "03/19/2025 13:30:29", "content": "

Hi everybody
\nthis is my solution Notebook
\nI have no idea why I got submission scoring error

", "rawMarkdown": "Hi everybody \nthis is my solution [Notebook](http://kaggle.com/code/arminajdehnia/birdclef-epaca-tdnn-inference)\nI have no idea why I got submission scoring error", "votes": null }, { "id": "3154190", "postDate": "03/19/2025 16:12:36", "content": "

Make sure you are not attempting to load any model pretrained weights with internet off.

\n

To debug, turn internet off on your notebook and point the input directory from test to train and see if it breaks anywhere. Try to simulate the inference pipeline on the test set as much as possible and see where it breaks.

", "rawMarkdown": "Make sure you are not attempting to load any model pretrained weights with internet off. \n\nTo debug, turn internet off on your notebook and point the input directory from test to train and see if it breaks anywhere. Try to simulate the inference pipeline on the test set as much as possible and see where it breaks.", "votes": null }, { "id": "3154196", "postDate": "03/19/2025 16:18:38", "content": "

\"I checked again and confirmed that internet access was off.

\n

Although test_soundscapes was empty for debugging, I used train_soundscapes for testing. Everything worked fine, and the submission file was successfully created with the required columns in the correct order.\"

\n

Could you take a look at the Nookbook when you get a chance?

", "rawMarkdown": "\"I checked again and confirmed that internet access was off.\n\nAlthough test_soundscapes was empty for debugging, I used train_soundscapes for testing. Everything worked fine, and the submission file was successfully created with the required columns in the correct order.\"\n\nCould you take a look at the Nookbook when you get a chance?", "votes": null }, { "id": "3154202", "postDate": "03/19/2025 16:36:20", "content": "

We strongly recommend adapting directly from the sample submission to minimize time lost to wrestling with file layouts, etc.

", "rawMarkdown": "We strongly recommend adapting directly from the [sample submission](https://www.kaggle.com/code/stefankahl/birdclef-2025-sample-submission) to minimize time lost to wrestling with file layouts, etc.", "votes": null }, { "id": "3154211", "postDate": "03/19/2025 16:50:38", "content": "

row_id needs to be a column in your submission.csv

", "rawMarkdown": "`row_id` needs to be a column in your `submission.csv`", "votes": null }, { "id": "3154272", "postDate": "03/19/2025 18:33:45", "content": "

I followed the instructions correctly, but I encountered a notebook timeout error, even though it didn’t run for more than 30 minutes.

", "rawMarkdown": "I followed the instructions correctly, but I encountered a notebook timeout error, even though it didn’t run for more than 30 minutes.", "votes": null }, { "id": "3154274", "postDate": "03/19/2025 18:35:10", "content": "

Do you mind if I ask you to check the lates version of this notebook

", "rawMarkdown": "Do you mind if I ask you to check the lates version of this notebook [](https://www.kaggle.com/code/arminajdehnia/birdclef-epaca-tdnn-inference)", "votes": null }, { "id": "3154461", "postDate": "03/20/2025 01:15:03", "content": "

I think it's because you're using GPU acceleration, the competition rules say that it must be run on CPU and if it's GPU, it'll timeout after 1 minute.

", "rawMarkdown": "I think it's because you're using GPU acceleration, the competition rules say that it must be run on CPU and if it's GPU, it'll timeout after 1 minute.", "votes": null }, { "id": "3154755", "postDate": "03/20/2025 10:55:48", "content": "

Thanks
\nI switched to using CPU instead of GPU then Notebook timeout error got ok
\nBut the scoring error remains.

\n

I check the notebook with Sample submission and everythings looks good. @tomdenton

", "rawMarkdown": "Thanks\nI switched to using CPU instead of GPU then Notebook timeout error got ok\nBut the scoring error remains.\n\nI check the notebook with [Sample submission](https://www.kaggle.com/code/stefankahl/birdclef-2025-sample-submission) and everythings looks good. @tomdenton", "votes": null }, { "id": "3154756", "postDate": "03/20/2025 10:56:46", "content": "

I have considered it but error still remains in last version.

", "rawMarkdown": "I have considered it but error still remains in last version.", "votes": null }, { "id": "3155111", "postDate": "03/20/2025 18:14:07", "content": "

I'm facing the same issue, my notebook runs fast since I'm uploading the weights of the model that I trained locally.
\nEven following the same code for the submission generation as the sample submission file, it fails at the scoring phase.

", "rawMarkdown": "I'm facing the same issue, my notebook runs fast since I'm uploading the weights of the model that I trained locally.\nEven following the same code for the submission generation as the sample submission file, it fails at the scoring phase.", "votes": null }, { "id": "3155678", "postDate": "03/21/2025 09:41:41", "content": "

Did you fix it?
\nif yes could you mention how do you do it?

", "rawMarkdown": "Did you fix it?\nif yes could you mention how do you do it?", "votes": null }, { "id": "3163158", "postDate": "03/30/2025 11:17:09", "content": "

Hello. Also got stuck with \"Submission Scoring Error\". Absolutely unclear what is the issue. No GPU, no internet access. Tried:

\n\n

Ran locally with one file soundscape_12345.ogg in directory. Columns are the same as in sample submission. Could you explain what is wrong with output?

", "rawMarkdown": "Hello. Also got stuck with \"Submission Scoring Error\". Absolutely unclear what is the issue. No GPU, no internet access. Tried:\n- load columns from sample submission csv\n- load columns from train dir\n- have seen sample workbook\n\nRan locally with one file soundscape_12345.ogg in directory. Columns are the same as in sample submission. Could you explain what is wrong with output?", "votes": null }, { "id": "3163221", "postDate": "03/30/2025 12:50:16", "content": "

The issue was with using the torchaudio.load function. I replaced it with librosa.load, and it worked correctly.
\nThe main problem was splitting the audio into 5-second windows.

", "rawMarkdown": "The issue was with using the torchaudio.load function. I replaced it with librosa.load, and it worked correctly.\nThe main problem was splitting the audio into 5-second windows.", "votes": null }, { "id": "3163238", "postDate": "03/30/2025 13:18:23", "content": "

I'm already using:
\ny, sr = librosa.load(path_to_file, sr=None)

\n

My guess was that column values don't sum to 1.0. But when I put 1.0 to one column result is the same. Skipping empty rows also doesn't help. Why just not to give a more explicit error?! This is really frustrating.

", "rawMarkdown": "I'm already using:\ny, sr = librosa.load(path_to_file, sr=None)\n\nMy guess was that column values don't sum to 1.0. But when I put 1.0 to one column result is the same. Skipping empty rows also doesn't help. Why just not to give a more explicit error?! This is really frustrating.", "votes": null } ], "comments": [ { "id": 3154190, "author_name": "connorjd", "author_url": "", "post_date": "03/19/2025 16:12:36", "content": "

Make sure you are not attempting to load any model pretrained weights with internet off.

\n

To debug, turn internet off on your notebook and point the input directory from test to train and see if it breaks anywhere. Try to simulate the inference pipeline on the test set as much as possible and see where it breaks.

", "votes": null, "replies": [ { "id": 3154196, "author_name": "arminajdehnia", "author_url": "", "post_date": "03/19/2025 16:18:38", "content": "

\"I checked again and confirmed that internet access was off.

\n

Although test_soundscapes was empty for debugging, I used train_soundscapes for testing. Everything worked fine, and the submission file was successfully created with the required columns in the correct order.\"

\n

Could you take a look at the Nookbook when you get a chance?

", "votes": null, "replies": [] } ] }, { "id": 3154202, "author_name": "tomdenton", "author_url": "", "post_date": "03/19/2025 16:36:20", "content": "

We strongly recommend adapting directly from the sample submission to minimize time lost to wrestling with file layouts, etc.

", "votes": null, "replies": [ { "id": 3154272, "author_name": "arminajdehnia", "author_url": "", "post_date": "03/19/2025 18:33:45", "content": "

I followed the instructions correctly, but I encountered a notebook timeout error, even though it didn’t run for more than 30 minutes.

", "votes": null, "replies": [] }, { "id": 3154274, "author_name": "arminajdehnia", "author_url": "", "post_date": "03/19/2025 18:35:10", "content": "

Do you mind if I ask you to check the lates version of this notebook

", "votes": null, "replies": [] } ] }, { "id": 3154211, "author_name": "mccocoful", "author_url": "", "post_date": "03/19/2025 16:50:38", "content": "

row_id needs to be a column in your submission.csv

", "votes": null, "replies": [ { "id": 3154756, "author_name": "arminajdehnia", "author_url": "", "post_date": "03/20/2025 10:56:46", "content": "

I have considered it but error still remains in last version.

", "votes": null, "replies": [] } ] }, { "id": 3154461, "author_name": "ricardoriveroh", "author_url": "", "post_date": "03/20/2025 01:15:03", "content": "

I think it's because you're using GPU acceleration, the competition rules say that it must be run on CPU and if it's GPU, it'll timeout after 1 minute.

", "votes": null, "replies": [ { "id": 3154755, "author_name": "arminajdehnia", "author_url": "", "post_date": "03/20/2025 10:55:48", "content": "

Thanks
\nI switched to using CPU instead of GPU then Notebook timeout error got ok
\nBut the scoring error remains.

\n

I check the notebook with Sample submission and everythings looks good. @tomdenton

", "votes": null, "replies": [ { "id": 3155111, "author_name": "ricardoriveroh", "author_url": "", "post_date": "03/20/2025 18:14:07", "content": "

I'm facing the same issue, my notebook runs fast since I'm uploading the weights of the model that I trained locally.
\nEven following the same code for the submission generation as the sample submission file, it fails at the scoring phase.

", "votes": null, "replies": [ { "id": 3155678, "author_name": "arminajdehnia", "author_url": "", "post_date": "03/21/2025 09:41:41", "content": "

Did you fix it?
\nif yes could you mention how do you do it?

", "votes": null, "replies": [] } ] } ] } ] }, { "id": 3163158, "author_name": "zharkomi", "author_url": "", "post_date": "03/30/2025 11:17:09", "content": "

Hello. Also got stuck with \"Submission Scoring Error\". Absolutely unclear what is the issue. No GPU, no internet access. Tried:

\n\n

Ran locally with one file soundscape_12345.ogg in directory. Columns are the same as in sample submission. Could you explain what is wrong with output?

", "votes": null, "replies": [ { "id": 3163221, "author_name": "arminajdehnia", "author_url": "", "post_date": "03/30/2025 12:50:16", "content": "

The issue was with using the torchaudio.load function. I replaced it with librosa.load, and it worked correctly.
\nThe main problem was splitting the audio into 5-second windows.

", "votes": null, "replies": [ { "id": 3163238, "author_name": "zharkomi", "author_url": "", "post_date": "03/30/2025 13:18:23", "content": "

I'm already using:
\ny, sr = librosa.load(path_to_file, sr=None)

\n

My guess was that column values don't sum to 1.0. But when I put 1.0 to one column result is the same. Skipping empty rows also doesn't help. Why just not to give a more explicit error?! This is really frustrating.

", "votes": null, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "3154069": "Hi everybody \nthis is my solution [Notebook](http://kaggle.com/code/arminajdehnia/birdclef-epaca-tdnn-inference)\nI have no idea why I got submission scoring error", "3154190": "Make sure you are not attempting to load any model pretrained weights with internet off. \n\nTo debug, turn internet off on your notebook and point the input directory from test to train and see if it breaks anywhere. Try to simulate the inference pipeline on the test set as much as possible and see where it breaks.", "3154196": "\"I checked again and confirmed that internet access was off.\n\nAlthough test_soundscapes was empty for debugging, I used train_soundscapes for testing. Everything worked fine, and the submission file was successfully created with the required columns in the correct order.\"\n\nCould you take a look at the Nookbook when you get a chance?", "3154202": "We strongly recommend adapting directly from the [sample submission](https://www.kaggle.com/code/stefankahl/birdclef-2025-sample-submission) to minimize time lost to wrestling with file layouts, etc.", "3154211": "`row_id` needs to be a column in your `submission.csv`", "3154272": "I followed the instructions correctly, but I encountered a notebook timeout error, even though it didn’t run for more than 30 minutes.", "3154274": "Do you mind if I ask you to check the lates version of this notebook [](https://www.kaggle.com/code/arminajdehnia/birdclef-epaca-tdnn-inference)", "3154461": "I think it's because you're using GPU acceleration, the competition rules say that it must be run on CPU and if it's GPU, it'll timeout after 1 minute.", "3154755": "Thanks\nI switched to using CPU instead of GPU then Notebook timeout error got ok\nBut the scoring error remains.\n\nI check the notebook with [Sample submission](https://www.kaggle.com/code/stefankahl/birdclef-2025-sample-submission) and everythings looks good. @tomdenton", "3154756": "I have considered it but error still remains in last version.", "3155111": "I'm facing the same issue, my notebook runs fast since I'm uploading the weights of the model that I trained locally.\nEven following the same code for the submission generation as the sample submission file, it fails at the scoring phase.", "3155678": "Did you fix it?\nif yes could you mention how do you do it?", "3163158": "Hello. Also got stuck with \"Submission Scoring Error\". Absolutely unclear what is the issue. No GPU, no internet access. Tried:\n- load columns from sample submission csv\n- load columns from train dir\n- have seen sample workbook\n\nRan locally with one file soundscape_12345.ogg in directory. Columns are the same as in sample submission. Could you explain what is wrong with output?", "3163221": "The issue was with using the torchaudio.load function. I replaced it with librosa.load, and it worked correctly.\nThe main problem was splitting the audio into 5-second windows.", "3163238": "I'm already using:\ny, sr = librosa.load(path_to_file, sr=None)\n\nMy guess was that column values don't sum to 1.0. But when I put 1.0 to one column result is the same. Skipping empty rows also doesn't help. Why just not to give a more explicit error?! This is really frustrating." }, "source": "meta" }