It runs fine on local test, but after submission I get Notebook Threw Exception. I am new to Kaggle and don't know how to resolve this. I would appreciate it if someone could tell me how to solve this problem or if I am writing the code incorrectly.
\nhttps://www.kaggle.com/code/naitoyuta/bfrb-ny-notebook
", "messages": [ { "id": "3232344", "postDate": "06/25/2025 15:39:59", "content": "It runs fine on local test, but after submission I get Notebook Threw Exception. I am new to Kaggle and don't know how to resolve this. I would appreciate it if someone could tell me how to solve this problem or if I am writing the code incorrectly.
\nhttps://www.kaggle.com/code/naitoyuta/bfrb-ny-notebook
", "rawMarkdown": "It runs fine on local test, but after submission I get Notebook Threw Exception. I am new to Kaggle and don't know how to resolve this. I would appreciate it if someone could tell me how to solve this problem or if I am writing the code incorrectly.\n\nhttps://www.kaggle.com/code/naitoyuta/bfrb-ny-notebook", "votes": null }, { "id": "3232361", "postDate": "06/25/2025 16:00:36", "content": "Are you using two branches? like one for imu and one for a model using all sensors. If you have a selection mechanism in your predict function to select which one of the two models you would use based on the data availability, it might be due to your imu model is providing wrong output. All the samples in the test.csv have all the data, it will not test your imu_model.
", "rawMarkdown": "Are you using two branches? like one for imu and one for a model using all sensors. If you have a selection mechanism in your predict function to select which one of the two models you would use based on the data availability, it might be due to your imu model is providing wrong output. All the samples in the test.csv have all the data, it will not test your imu_model.", "votes": null }, { "id": "3232364", "postDate": "06/25/2025 16:03:28", "content": "Even if you use just one model, you need to be aware that half of the evaluation will be done only with imu sensor data, so you need to check if you handle the null values correctly for those sequences that have only imu sensor data (at this time other columns are null values)
", "rawMarkdown": "Even if you use just one model, you need to be aware that half of the evaluation will be done only with imu sensor data, so you need to check if you handle the null values correctly for those sequences that have only imu sensor data (at this time other columns are null values)", "votes": null }, { "id": "3233285", "postDate": "06/26/2025 16:41:30", "content": "Thank you very much! As you pointed out, my process did not properly handle null values when the data was one data for IMU only.I was able to improve and submit the data successfully.The score is very bad though.
", "rawMarkdown": "Thank you very much! As you pointed out, my process did not properly handle null values when the data was one data for IMU only.I was able to improve and submit the data successfully.The score is very bad though.", "votes": null } ], "comments": [ { "id": 3232361, "author_name": "georgewang73", "author_url": "", "post_date": "06/25/2025 16:00:36", "content": "Are you using two branches? like one for imu and one for a model using all sensors. If you have a selection mechanism in your predict function to select which one of the two models you would use based on the data availability, it might be due to your imu model is providing wrong output. All the samples in the test.csv have all the data, it will not test your imu_model.
", "votes": null, "replies": [ { "id": 3232364, "author_name": "georgewang73", "author_url": "", "post_date": "06/25/2025 16:03:28", "content": "Even if you use just one model, you need to be aware that half of the evaluation will be done only with imu sensor data, so you need to check if you handle the null values correctly for those sequences that have only imu sensor data (at this time other columns are null values)
", "votes": null, "replies": [ { "id": 3233285, "author_name": "naitoyuta", "author_url": "", "post_date": "06/26/2025 16:41:30", "content": "Thank you very much! As you pointed out, my process did not properly handle null values when the data was one data for IMU only.I was able to improve and submit the data successfully.The score is very bad though.
", "votes": null, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "3232344": "It runs fine on local test, but after submission I get Notebook Threw Exception. I am new to Kaggle and don't know how to resolve this. I would appreciate it if someone could tell me how to solve this problem or if I am writing the code incorrectly.\n\nhttps://www.kaggle.com/code/naitoyuta/bfrb-ny-notebook", "3232361": "Are you using two branches? like one for imu and one for a model using all sensors. If you have a selection mechanism in your predict function to select which one of the two models you would use based on the data availability, it might be due to your imu model is providing wrong output. All the samples in the test.csv have all the data, it will not test your imu_model.", "3232364": "Even if you use just one model, you need to be aware that half of the evaluation will be done only with imu sensor data, so you need to check if you handle the null values correctly for those sequences that have only imu sensor data (at this time other columns are null values)", "3233285": "Thank you very much! As you pointed out, my process did not properly handle null values when the data was one data for IMU only.I was able to improve and submit the data successfully.The score is very bad though." }, "source": "meta" }