Hi everyone! We got some weird issue with our submissions: firstly we submitted poorly trained model, to check our submission pipeline. Everything was good and submission was accepted. Then we trained the same model but on a larger number of epochs. Submission code and everything's the same, model's publicly available, but we're having \"Notebook Timeout\" exception withh well trained model. While submitting our prediction we're only
\n importing libraries (without downloading anything), loading model checkpoint, predicting files from \"test_soundscapes\" and exporting as submission.csv file. Guys, did anyone have the same problem?
Hi everyone! We got some weird issue with our submissions: firstly we submitted poorly trained model, to check our submission pipeline. Everything was good and submission was accepted. Then we trained the same model but on a larger number of epochs. Submission code and everything's the same, model's publicly available, but we're having \"Notebook Timeout\" exception withh well trained model. While submitting our prediction we're only
\n importing libraries (without downloading anything), loading model checkpoint, predicting files from \"test_soundscapes\" and exporting as submission.csv file. Guys, did anyone have the same problem?
FYI. As a rule of thumb, if you make the same submission, it is possible that only one of the submissions will time out.
\nThis happens due to slight differences in the HW specs of the assigned inference server.
Therefore, if you are submitting code close to the submission requirements (in this competition, CPU Notebook <= 90 minutes run-time), you may be able to submit multiple times and still be scored without timing out.
", "rawMarkdown": "FYI. As a rule of thumb, if you make the same submission, it is possible that only one of the submissions will time out.\nThis happens due to slight differences in the HW specs of the assigned inference server.\n\nTherefore, if you are submitting code close to the submission requirements (in this competition, CPU Notebook <= 90 minutes run-time), you may be able to submit multiple times and still be scored without timing out.", "votes": null }, { "id": "3211273", "postDate": "05/28/2025 08:45:01", "content": "Thank you for your answer! I got you, but I think there's something else: Our poorly trained model was sucessfully submitted 2 times, each submission took 12 minutes, but well trained model was never accepted. We've tried it for about 3-4 times with a new one, so I think there's another problem.
", "rawMarkdown": "Thank you for your answer! I got you, but I think there's something else: Our poorly trained model was sucessfully submitted 2 times, each submission took 12 minutes, but well trained model was never accepted. We've tried it for about 3-4 times with a new one, so I think there's another problem.", "votes": null } ], "comments": [ { "id": 3211268, "author_name": "hideyukizushi", "author_url": "", "post_date": "05/28/2025 08:36:05", "content": "FYI. As a rule of thumb, if you make the same submission, it is possible that only one of the submissions will time out.
\nThis happens due to slight differences in the HW specs of the assigned inference server.
Therefore, if you are submitting code close to the submission requirements (in this competition, CPU Notebook <= 90 minutes run-time), you may be able to submit multiple times and still be scored without timing out.
", "votes": null, "replies": [ { "id": 3211273, "author_name": "darkl1ght", "author_url": "", "post_date": "05/28/2025 08:45:01", "content": "Thank you for your answer! I got you, but I think there's something else: Our poorly trained model was sucessfully submitted 2 times, each submission took 12 minutes, but well trained model was never accepted. We've tried it for about 3-4 times with a new one, so I think there's another problem.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "3211190": "Hi everyone! We got some weird issue with our submissions: firstly we submitted poorly trained model, to check our submission pipeline. Everything was good and submission was accepted. Then we trained the same model but on a larger number of epochs. Submission code and **everything's the same**, model's publicly available, but we're having \"Notebook Timeout\" exception withh well trained model. While submitting our prediction we're only \n importing libraries (without downloading anything), loading model checkpoint, predicting files from \"test_soundscapes\" and exporting as submission.csv file. Guys, did anyone have the same problem?", "3211268": "FYI. As a rule of thumb, if you make the same submission, it is possible that only one of the submissions will time out.\nThis happens due to slight differences in the HW specs of the assigned inference server.\n\nTherefore, if you are submitting code close to the submission requirements (in this competition, CPU Notebook <= 90 minutes run-time), you may be able to submit multiple times and still be scored without timing out.", "3211273": "Thank you for your answer! I got you, but I think there's something else: Our poorly trained model was sucessfully submitted 2 times, each submission took 12 minutes, but well trained model was never accepted. We've tried it for about 3-4 times with a new one, so I think there's another problem." }, "source": "meta" }