{ "id": 567960, "title": "Submission File Format Issue: Hidden Test Set Not Being Populated?", "url": "/competitions/birdclef-2025/discussion/567960", "author_name": "", "post_date": "2025-03-13T02:33:58.784941200Z", "votes": null, "comment_count": 20, "views": 0, "content": "

Hi everyone,

\n

I'm encountering an issue with my submission for the BirdCLEF+ 2025 competition. When I run my notebook in the submission environment, my code checks the test folder (../input/birdclef-2025/test_soundscapes) for audio files. However, the folder only contains a readme.txt file, and the code detects 0 test files. As a result, my notebook falls back to generating a dummy submission (by loading the sample_submission.csv) and creates a submission file with a different row count and structure than expected.

\n

This leads to a \"Submission Scoring Error\" indicating that the generated submission file has an incorrect format (e.g., wrong number of rows or columns, empty values, etc.).

\n

I understand that Kaggle's system hides the actual test files to prevent overfitting and that the hidden test set is injected during the final evaluation run. However, my concern is that, in my submission run, the hidden test set doesn't seem to be populated at all, and I'm forced to generate a dummy submission, which then doesn't match the expected format.

\n

Has anyone experienced similar issues where the hidden test set wasn't visible or populated during submission? Any advice or solutions would be greatly appreciated.

\n

Thanks in advance!

", "messages": [ { "id": "3148329", "postDate": "03/13/2025 02:33:58", "content": "

Hi everyone,

\n

I'm encountering an issue with my submission for the BirdCLEF+ 2025 competition. When I run my notebook in the submission environment, my code checks the test folder (../input/birdclef-2025/test_soundscapes) for audio files. However, the folder only contains a readme.txt file, and the code detects 0 test files. As a result, my notebook falls back to generating a dummy submission (by loading the sample_submission.csv) and creates a submission file with a different row count and structure than expected.

\n

This leads to a \"Submission Scoring Error\" indicating that the generated submission file has an incorrect format (e.g., wrong number of rows or columns, empty values, etc.).

\n

I understand that Kaggle's system hides the actual test files to prevent overfitting and that the hidden test set is injected during the final evaluation run. However, my concern is that, in my submission run, the hidden test set doesn't seem to be populated at all, and I'm forced to generate a dummy submission, which then doesn't match the expected format.

\n

Has anyone experienced similar issues where the hidden test set wasn't visible or populated during submission? Any advice or solutions would be greatly appreciated.

\n

Thanks in advance!

", "rawMarkdown": "Hi everyone,\n\nI'm encountering an issue with my submission for the BirdCLEF+ 2025 competition. When I run my notebook in the submission environment, my code checks the test folder `(../input/birdclef-2025/test_soundscapes)` for audio files. However, the folder only contains a readme.txt file, and the code detects 0 test files. As a result, my notebook falls back to generating a dummy submission (by loading the sample_submission.csv) and creates a submission file with a different row count and structure than expected.\n\nThis leads to a \"Submission Scoring Error\" indicating that the generated submission file has an incorrect format (e.g., wrong number of rows or columns, empty values, etc.).\n\nI understand that Kaggle's system hides the actual test files to prevent overfitting and that the hidden test set is injected during the final evaluation run. However, my concern is that, in my submission run, the hidden test set doesn't seem to be populated at all, and I'm forced to generate a dummy submission, which then doesn't match the expected format.\n\nHas anyone experienced similar issues where the hidden test set wasn't visible or populated during submission? Any advice or solutions would be greatly appreciated.\n\nThanks in advance!", "votes": null }, { "id": "3148389", "postDate": "03/13/2025 04:02:06", "content": "

HI, Hardi -
\nIt can be tricky to get the file structure right. We strongly recommend adapting an existing notebook, like the sample submission, to ensure you access the files correctly.

\n

https://www.kaggle.com/code/stefankahl/birdclef-2025-sample-submission

", "rawMarkdown": "HI, Hardi -\nIt can be tricky to get the file structure right. We strongly recommend adapting an existing notebook, like the sample submission, to ensure you access the files correctly.\n\nhttps://www.kaggle.com/code/stefankahl/birdclef-2025-sample-submission", "votes": null }, { "id": "3182095", "postDate": "04/18/2025 18:15:34", "content": "

I'm having the same issue.

", "rawMarkdown": "I'm having the same issue.", "votes": null }, { "id": "3186633", "postDate": "04/25/2025 01:40:58", "content": "

Were you able to resolve the submission issue? @hardikim

", "rawMarkdown": "Were you able to resolve the submission issue? @hardikim", "votes": null }, { "id": "3187394", "postDate": "04/26/2025 00:29:32", "content": "

I'm running with the correct restrictions but no test soundscapes are populated. Are we allowed to install dependencies?

", "rawMarkdown": "I'm running with the correct restrictions but no test soundscapes are populated. Are we allowed to install dependencies?", "votes": null }, { "id": "3187853", "postDate": "04/26/2025 16:16:04", "content": "

You will never see the test folder populated. You have to turn off internet in the notebook and then submit it.

\n

The best practice from what I saw is to make the notebook to run in two modes: you check if test folder has any ogg files and if not, then run the notebook in second mode ('train') and grab files from train_soundscapes (like f'{mode}_soundscapes').

\n

That way notebook won't crash when you save it and produce a valid submission.csv anyway (with examples).

\n

The other thing I found is that you need to generate row_id very carefully. If you mess up final second of the sample being scores, you still get the error OP struggles with.

", "rawMarkdown": "You will never see the test folder populated. You have to turn off internet in the notebook and then submit it.\n\nThe best practice from what I saw is to make the notebook to run in two modes: you check if test folder has any ogg files and if not, then run the notebook in second mode ('train') and grab files from train_soundscapes (like f'{mode}_soundscapes').\n\nThat way notebook won't crash when you save it and produce a valid submission.csv anyway (with examples).\n\nThe other thing I found is that you need to generate row_id very carefully. If you mess up final second of the sample being scores, you still get the error OP struggles with.", "votes": null }, { "id": "3187975", "postDate": "04/26/2025 19:42:37", "content": "

The notebook runs twice, the logs etc is before it runs, then it populates and runs. The way I knew it was working was the results actually ran with the csv.

", "rawMarkdown": "The notebook runs twice, the logs etc is before it runs, then it populates and runs. The way I knew it was working was the results actually ran with the csv.", "votes": null }, { "id": "3195356", "postDate": "05/07/2025 00:06:50", "content": "

Having this issue as well

", "rawMarkdown": "Having this issue as well", "votes": null }, { "id": "3203349", "postDate": "05/16/2025 15:44:07", "content": "

Hi Tom,
\nThank you for the suggestion. I'm having the same issue as Hardi, and using the same method for retrieving file paths shown in the notebook that you referenced does not solve the issue. As a debugging step I printed the content of the folder but the only path was \"readme.txt\". I will keep looking for a solution, but just thought I should share that the problem may not be with getting the file structure right.

", "rawMarkdown": "Hi Tom,\nThank you for the suggestion. I'm having the same issue as Hardi, and using the same method for retrieving file paths shown in the notebook that you referenced does not solve the issue. As a debugging step I printed the content of the folder but the only path was \"readme.txt\". I will keep looking for a solution, but just thought I should share that the problem may not be with getting the file structure right.", "votes": null }, { "id": "3203367", "postDate": "05/16/2025 16:11:14", "content": "

Hi, Masato -
\nThe test files are only visible to the notebook during the submission run.

", "rawMarkdown": "Hi, Masato -\nThe test files are only visible to the notebook during the submission run.", "votes": null }, { "id": "3203401", "postDate": "05/16/2025 17:20:04", "content": "

Tom, thanks for replying so soon. This is embarrassing to admit, but I just learned the difference between saving & commiting vs. submission in competitions. I thought committing the code would have the same effect, but obviously I was wrong. Thank you for helping me reach this point.

", "rawMarkdown": "Tom, thanks for replying so soon. This is embarrassing to admit, but I just learned the difference between saving & commiting vs. submission in competitions. I thought committing the code would have the same effect, but obviously I was wrong. Thank you for helping me reach this point.", "votes": null }, { "id": "3209712", "postDate": "05/26/2025 07:47:33", "content": "

I have also encountered related problems. I know that test_soundscapes is hidden when notebook is running. But when I submitted it, I detected that there was no ogg file in the directory of test_soundscapes, which made it impossible for me to submit the results. It would be a great honor if someone could guide my code (https://www.kaggle.com/code/dairuiwei/birdclef-2025-with-drw-second).

", "rawMarkdown": "I have also encountered related problems. I know that test_soundscapes is hidden when notebook is running. But when I submitted it, I detected that there was no ogg file in the directory of test_soundscapes, which made it impossible for me to submit the results. It would be a great honor if someone could guide my code (https://www.kaggle.com/code/dairuiwei/birdclef-2025-with-drw-second).", "votes": null }, { "id": "3209768", "postDate": "05/26/2025 09:35:06", "content": "

Your code is private apparently. I couldn't open the link!

\n

Also i wonder how did you detected it, since we are not allowed to peek into submission and have a limited knowledge of what is going on inside there.

\n

Maybe you mean you ran your notebook? It's a different thing!

", "rawMarkdown": "Your code is private apparently. I couldn't open the link!\n\nAlso i wonder how did you detected it, since we are not allowed to peek into submission and have a limited knowledge of what is going on inside there.\n\nMaybe you mean you ran your notebook? It's a different thing!", "votes": null }, { "id": "3209786", "postDate": "05/26/2025 10:00:00", "content": "

Thank you for your reply! You can click on my avatar to view the code now.
\nRegarding the detection of test_soundscapes, I added some checking statements in my code. These statements are designed to output whether there are files in the test_soundscapes directory when I submit the notebook. I understand that we're not supposed to peek into the test set in an improper way, but these statements are just for basic verification during the submission process to help me figure out why I can't read the files.
\nIf you see my clumsy code, please don't laugh at me. 🥺
\nI would really appreciate it if you could take a look and offer any suggestions on what might be going wrong with the file - reading process.

", "rawMarkdown": "Thank you for your reply! You can click on my avatar to view the code now.\nRegarding the detection of test_soundscapes, I added some checking statements in my code. These statements are designed to output whether there are files in the test_soundscapes directory when I submit the notebook. I understand that we're not supposed to peek into the test set in an improper way, but these statements are just for basic verification during the submission process to help me figure out why I can't read the files.\nIf you see my clumsy code, please don't laugh at me. 🥺\nI would really appreciate it if you could take a look and offer any suggestions on what might be going wrong with the file - reading process.", "votes": null }, { "id": "3209814", "postDate": "05/26/2025 11:17:14", "content": "

I checked your code, the problem is that when you save your notebook (run and save) the test folder will be still empty. You need to turn off the internet in the notebook, generate a csv and click submit to competition. And only there the folder is populated.

\n

There are different work arounds here. The best practice imo is to use mode variable, check if test folder is empty and assign to it a 'test', if it is not, or 'train' if it is empty.

\n

That way you still may run your model on train_soundscapes and check if it doesn't collapse.

", "rawMarkdown": "I checked your code, the problem is that when you save your notebook (run and save) the test folder will be still empty. You need to turn off the internet in the notebook, generate a csv and click submit to competition. And only there the folder is populated.\n\nThere are different work arounds here. The best practice imo is to use mode variable, check if test folder is empty and assign to it a 'test', if it is not, or 'train' if it is empty.\n\nThat way you still may run your model on train_soundscapes and check if it doesn't collapse.", "votes": null }, { "id": "3209846", "postDate": "05/26/2025 12:01:51", "content": "

Thanks for your guidance. I actually checked the test and submission environments before. But I still gave your suggestion a try. Sadly, I still can't detect anything in test_soundscapes (😂). I've used all 5 submission attempts for today. Anyway, it's great to discuss code problems with you.

", "rawMarkdown": "Thanks for your guidance. I actually checked the test and submission environments before. But I still gave your suggestion a try. Sadly, I still can't detect anything in test_soundscapes (😂). I've used all 5 submission attempts for today. Anyway, it's great to discuss code problems with you.", "votes": null }, { "id": "3209908", "postDate": "05/26/2025 13:29:51", "content": "

Use this structure:

\n
mode = \nsoundscape_dir = \naudio_files = ([f  f  os.listdir(soundscape_dir)  f.endswith()])\n\n  audio_files:\n    mode = \n    soundscape_dir = \n    audio_files = (os.listdir(soundscape_dir))[:]\n
\n

And then use mode variable to create a structure for row_id you will insert into submission.csv. I am sure you'll figure that out.

", "rawMarkdown": "Use this structure:\n\n mode = 'test'\n soundscape_dir = f\"../input/birdclef-2025/{mode}_soundscapes\"\n audio_files = sorted([f for f in os.listdir(soundscape_dir) if f.endswith('.ogg')])\n # fallback to 'train' with few files\n if not audio_files:\n mode = 'train'\n soundscape_dir = f\"../input/birdclef-2025/{mode}_soundscapes\"\n audio_files = sorted(os.listdir(soundscape_dir))[:5]\n\nAnd then use mode variable to create a structure for row_id you will insert into submission.csv. I am sure you'll figure that out.", "votes": null }, { "id": "3209929", "postDate": "05/26/2025 14:10:18", "content": "

Thanks for the guidance! When following your advice to predict with train_soundscapes, I found a column mismatch issue. Fixed it and will submit the updated code tomorrow!

", "rawMarkdown": "Thanks for the guidance! When following your advice to predict with train_soundscapes, I found a column mismatch issue. Fixed it and will submit the updated code tomorrow!", "votes": null }, { "id": "3214864", "postDate": "06/01/2025 07:51:04", "content": "

Hi DRW, what was the column mismatch issue you found?

", "rawMarkdown": "Hi DRW, what was the column mismatch issue you found?", "votes": null }, { "id": "3214940", "postDate": "06/01/2025 10:04:27", "content": "

In fact, I found that there have already been posts discussing this issue: https://www.kaggle.com/competitions/birdclef-2025/discussion/567960

", "rawMarkdown": "In fact, I found that there have already been posts discussing this issue: https://www.kaggle.com/competitions/birdclef-2025/discussion/567960", "votes": null }, { "id": "3214943", "postDate": "06/01/2025 10:09:59", "content": "

You just need to create a dictionary containing row_id and all the category predicted values as the first column, and check whether the length of each predicted value is consistent with the number of category labels, and finally check whether it is missing and fill it in. That's all I did.

", "rawMarkdown": "You just need to create a dictionary containing row_id and all the category predicted values as the first column, and check whether the length of each predicted value is consistent with the number of category labels, and finally check whether it is missing and fill it in. That's all I did.", "votes": null } ], "comments": [ { "id": 3148389, "author_name": "tomdenton", "author_url": "", "post_date": "03/13/2025 04:02:06", "content": "

HI, Hardi -
\nIt can be tricky to get the file structure right. We strongly recommend adapting an existing notebook, like the sample submission, to ensure you access the files correctly.

\n

https://www.kaggle.com/code/stefankahl/birdclef-2025-sample-submission

", "votes": null, "replies": [ { "id": 3203349, "author_name": "masatokan", "author_url": "", "post_date": "05/16/2025 15:44:07", "content": "

Hi Tom,
\nThank you for the suggestion. I'm having the same issue as Hardi, and using the same method for retrieving file paths shown in the notebook that you referenced does not solve the issue. As a debugging step I printed the content of the folder but the only path was \"readme.txt\". I will keep looking for a solution, but just thought I should share that the problem may not be with getting the file structure right.

", "votes": null, "replies": [ { "id": 3203367, "author_name": "tomdenton", "author_url": "", "post_date": "05/16/2025 16:11:14", "content": "

Hi, Masato -
\nThe test files are only visible to the notebook during the submission run.

", "votes": null, "replies": [ { "id": 3203401, "author_name": "masatokan", "author_url": "", "post_date": "05/16/2025 17:20:04", "content": "

Tom, thanks for replying so soon. This is embarrassing to admit, but I just learned the difference between saving & commiting vs. submission in competitions. I thought committing the code would have the same effect, but obviously I was wrong. Thank you for helping me reach this point.

", "votes": null, "replies": [] } ] } ] } ] }, { "id": 3182095, "author_name": "patanet7", "author_url": "", "post_date": "04/18/2025 18:15:34", "content": "

I'm having the same issue.

", "votes": null, "replies": [] }, { "id": 3186633, "author_name": "varadrajbartakke919", "author_url": "", "post_date": "04/25/2025 01:40:58", "content": "

Were you able to resolve the submission issue? @hardikim

", "votes": null, "replies": [] }, { "id": 3187394, "author_name": "bgeier", "author_url": "", "post_date": "04/26/2025 00:29:32", "content": "

I'm running with the correct restrictions but no test soundscapes are populated. Are we allowed to install dependencies?

", "votes": null, "replies": [ { "id": 3187853, "author_name": "alexandergremyakov", "author_url": "", "post_date": "04/26/2025 16:16:04", "content": "

You will never see the test folder populated. You have to turn off internet in the notebook and then submit it.

\n

The best practice from what I saw is to make the notebook to run in two modes: you check if test folder has any ogg files and if not, then run the notebook in second mode ('train') and grab files from train_soundscapes (like f'{mode}_soundscapes').

\n

That way notebook won't crash when you save it and produce a valid submission.csv anyway (with examples).

\n

The other thing I found is that you need to generate row_id very carefully. If you mess up final second of the sample being scores, you still get the error OP struggles with.

", "votes": null, "replies": [] }, { "id": 3187975, "author_name": "patanet7", "author_url": "", "post_date": "04/26/2025 19:42:37", "content": "

The notebook runs twice, the logs etc is before it runs, then it populates and runs. The way I knew it was working was the results actually ran with the csv.

", "votes": null, "replies": [] } ] }, { "id": 3195356, "author_name": "benthecoder", "author_url": "", "post_date": "05/07/2025 00:06:50", "content": "

Having this issue as well

", "votes": null, "replies": [] }, { "id": 3209712, "author_name": "dairuiwei", "author_url": "", "post_date": "05/26/2025 07:47:33", "content": "

I have also encountered related problems. I know that test_soundscapes is hidden when notebook is running. But when I submitted it, I detected that there was no ogg file in the directory of test_soundscapes, which made it impossible for me to submit the results. It would be a great honor if someone could guide my code (https://www.kaggle.com/code/dairuiwei/birdclef-2025-with-drw-second).

", "votes": null, "replies": [ { "id": 3209768, "author_name": "alexandergremyakov", "author_url": "", "post_date": "05/26/2025 09:35:06", "content": "

Your code is private apparently. I couldn't open the link!

\n

Also i wonder how did you detected it, since we are not allowed to peek into submission and have a limited knowledge of what is going on inside there.

\n

Maybe you mean you ran your notebook? It's a different thing!

", "votes": null, "replies": [ { "id": 3209786, "author_name": "dairuiwei", "author_url": "", "post_date": "05/26/2025 10:00:00", "content": "

Thank you for your reply! You can click on my avatar to view the code now.
\nRegarding the detection of test_soundscapes, I added some checking statements in my code. These statements are designed to output whether there are files in the test_soundscapes directory when I submit the notebook. I understand that we're not supposed to peek into the test set in an improper way, but these statements are just for basic verification during the submission process to help me figure out why I can't read the files.
\nIf you see my clumsy code, please don't laugh at me. 🥺
\nI would really appreciate it if you could take a look and offer any suggestions on what might be going wrong with the file - reading process.

", "votes": null, "replies": [ { "id": 3209814, "author_name": "alexandergremyakov", "author_url": "", "post_date": "05/26/2025 11:17:14", "content": "

I checked your code, the problem is that when you save your notebook (run and save) the test folder will be still empty. You need to turn off the internet in the notebook, generate a csv and click submit to competition. And only there the folder is populated.

\n

There are different work arounds here. The best practice imo is to use mode variable, check if test folder is empty and assign to it a 'test', if it is not, or 'train' if it is empty.

\n

That way you still may run your model on train_soundscapes and check if it doesn't collapse.

", "votes": null, "replies": [ { "id": 3209846, "author_name": "dairuiwei", "author_url": "", "post_date": "05/26/2025 12:01:51", "content": "

Thanks for your guidance. I actually checked the test and submission environments before. But I still gave your suggestion a try. Sadly, I still can't detect anything in test_soundscapes (😂). I've used all 5 submission attempts for today. Anyway, it's great to discuss code problems with you.

", "votes": null, "replies": [ { "id": 3209908, "author_name": "alexandergremyakov", "author_url": "", "post_date": "05/26/2025 13:29:51", "content": "

Use this structure:

\n
mode = \nsoundscape_dir = \naudio_files = ([f  f  os.listdir(soundscape_dir)  f.endswith()])\n\n  audio_files:\n    mode = \n    soundscape_dir = \n    audio_files = (os.listdir(soundscape_dir))[:]\n
\n

And then use mode variable to create a structure for row_id you will insert into submission.csv. I am sure you'll figure that out.

", "votes": null, "replies": [ { "id": 3209929, "author_name": "dairuiwei", "author_url": "", "post_date": "05/26/2025 14:10:18", "content": "

Thanks for the guidance! When following your advice to predict with train_soundscapes, I found a column mismatch issue. Fixed it and will submit the updated code tomorrow!

", "votes": null, "replies": [ { "id": 3214864, "author_name": "a1dot1h", "author_url": "", "post_date": "06/01/2025 07:51:04", "content": "

Hi DRW, what was the column mismatch issue you found?

", "votes": null, "replies": [ { "id": 3214940, "author_name": "dairuiwei", "author_url": "", "post_date": "06/01/2025 10:04:27", "content": "

In fact, I found that there have already been posts discussing this issue: https://www.kaggle.com/competitions/birdclef-2025/discussion/567960

", "votes": null, "replies": [] }, { "id": 3214943, "author_name": "dairuiwei", "author_url": "", "post_date": "06/01/2025 10:09:59", "content": "

You just need to create a dictionary containing row_id and all the category predicted values as the first column, and check whether the length of each predicted value is consistent with the number of category labels, and finally check whether it is missing and fill it in. That's all I did.

", "votes": null, "replies": [] } ] } ] } ] } ] } ] } ] } ] } ] } ], "raw_markdown_by_id": { "3148329": "Hi everyone,\n\nI'm encountering an issue with my submission for the BirdCLEF+ 2025 competition. When I run my notebook in the submission environment, my code checks the test folder `(../input/birdclef-2025/test_soundscapes)` for audio files. However, the folder only contains a readme.txt file, and the code detects 0 test files. As a result, my notebook falls back to generating a dummy submission (by loading the sample_submission.csv) and creates a submission file with a different row count and structure than expected.\n\nThis leads to a \"Submission Scoring Error\" indicating that the generated submission file has an incorrect format (e.g., wrong number of rows or columns, empty values, etc.).\n\nI understand that Kaggle's system hides the actual test files to prevent overfitting and that the hidden test set is injected during the final evaluation run. However, my concern is that, in my submission run, the hidden test set doesn't seem to be populated at all, and I'm forced to generate a dummy submission, which then doesn't match the expected format.\n\nHas anyone experienced similar issues where the hidden test set wasn't visible or populated during submission? Any advice or solutions would be greatly appreciated.\n\nThanks in advance!", "3148389": "HI, Hardi -\nIt can be tricky to get the file structure right. We strongly recommend adapting an existing notebook, like the sample submission, to ensure you access the files correctly.\n\nhttps://www.kaggle.com/code/stefankahl/birdclef-2025-sample-submission", "3182095": "I'm having the same issue.", "3186633": "Were you able to resolve the submission issue? @hardikim", "3187394": "I'm running with the correct restrictions but no test soundscapes are populated. Are we allowed to install dependencies?", "3187853": "You will never see the test folder populated. You have to turn off internet in the notebook and then submit it.\n\nThe best practice from what I saw is to make the notebook to run in two modes: you check if test folder has any ogg files and if not, then run the notebook in second mode ('train') and grab files from train_soundscapes (like f'{mode}_soundscapes').\n\nThat way notebook won't crash when you save it and produce a valid submission.csv anyway (with examples).\n\nThe other thing I found is that you need to generate row_id very carefully. If you mess up final second of the sample being scores, you still get the error OP struggles with.", "3187975": "The notebook runs twice, the logs etc is before it runs, then it populates and runs. The way I knew it was working was the results actually ran with the csv.", "3195356": "Having this issue as well", "3203349": "Hi Tom,\nThank you for the suggestion. I'm having the same issue as Hardi, and using the same method for retrieving file paths shown in the notebook that you referenced does not solve the issue. As a debugging step I printed the content of the folder but the only path was \"readme.txt\". I will keep looking for a solution, but just thought I should share that the problem may not be with getting the file structure right.", "3203367": "Hi, Masato -\nThe test files are only visible to the notebook during the submission run.", "3203401": "Tom, thanks for replying so soon. This is embarrassing to admit, but I just learned the difference between saving & commiting vs. submission in competitions. I thought committing the code would have the same effect, but obviously I was wrong. Thank you for helping me reach this point.", "3209712": "I have also encountered related problems. I know that test_soundscapes is hidden when notebook is running. But when I submitted it, I detected that there was no ogg file in the directory of test_soundscapes, which made it impossible for me to submit the results. It would be a great honor if someone could guide my code (https://www.kaggle.com/code/dairuiwei/birdclef-2025-with-drw-second).", "3209768": "Your code is private apparently. I couldn't open the link!\n\nAlso i wonder how did you detected it, since we are not allowed to peek into submission and have a limited knowledge of what is going on inside there.\n\nMaybe you mean you ran your notebook? It's a different thing!", "3209786": "Thank you for your reply! You can click on my avatar to view the code now.\nRegarding the detection of test_soundscapes, I added some checking statements in my code. These statements are designed to output whether there are files in the test_soundscapes directory when I submit the notebook. I understand that we're not supposed to peek into the test set in an improper way, but these statements are just for basic verification during the submission process to help me figure out why I can't read the files.\nIf you see my clumsy code, please don't laugh at me. 🥺\nI would really appreciate it if you could take a look and offer any suggestions on what might be going wrong with the file - reading process.", "3209814": "I checked your code, the problem is that when you save your notebook (run and save) the test folder will be still empty. You need to turn off the internet in the notebook, generate a csv and click submit to competition. And only there the folder is populated.\n\nThere are different work arounds here. The best practice imo is to use mode variable, check if test folder is empty and assign to it a 'test', if it is not, or 'train' if it is empty.\n\nThat way you still may run your model on train_soundscapes and check if it doesn't collapse.", "3209846": "Thanks for your guidance. I actually checked the test and submission environments before. But I still gave your suggestion a try. Sadly, I still can't detect anything in test_soundscapes (😂). I've used all 5 submission attempts for today. Anyway, it's great to discuss code problems with you.", "3209908": "Use this structure:\n\n mode = 'test'\n soundscape_dir = f\"../input/birdclef-2025/{mode}_soundscapes\"\n audio_files = sorted([f for f in os.listdir(soundscape_dir) if f.endswith('.ogg')])\n # fallback to 'train' with few files\n if not audio_files:\n mode = 'train'\n soundscape_dir = f\"../input/birdclef-2025/{mode}_soundscapes\"\n audio_files = sorted(os.listdir(soundscape_dir))[:5]\n\nAnd then use mode variable to create a structure for row_id you will insert into submission.csv. I am sure you'll figure that out.", "3209929": "Thanks for the guidance! When following your advice to predict with train_soundscapes, I found a column mismatch issue. Fixed it and will submit the updated code tomorrow!", "3214864": "Hi DRW, what was the column mismatch issue you found?", "3214940": "In fact, I found that there have already been posts discussing this issue: https://www.kaggle.com/competitions/birdclef-2025/discussion/567960", "3214943": "You just need to create a dictionary containing row_id and all the category predicted values as the first column, and check whether the length of each predicted value is consistent with the number of category labels, and finally check whether it is missing and fill it in. That's all I did." }, "source": "meta" }