Hello,
\nAll these days, we used to test our networks internally and upload the predictions to obtain the AUC value.
\nNow we have put together all our processing pieces and made it a pipeline to submit for the final testing phase.
\nWhen we run this pipeline on a subsample of the validation data, it runs successfully. However, when we run it on all the cases, we are facing errors.
The error we get is \"Notebook Threw Exception\".
\nThe description of the error as on the website is \"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\".
Can someone please help us in this regard ?
", "messages": [ { "id": "1544855", "postDate": "10/14/2021 18:01:39", "content": "Hello,
\nAll these days, we used to test our networks internally and upload the predictions to obtain the AUC value.
\nNow we have put together all our processing pieces and made it a pipeline to submit for the final testing phase.
\nWhen we run this pipeline on a subsample of the validation data, it runs successfully. However, when we run it on all the cases, we are facing errors.
The error we get is \"Notebook Threw Exception\".
\nThe description of the error as on the website is \"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\".
Can someone please help us in this regard ?
", "rawMarkdown": "Hello,\n\nAll these days, we used to test our networks internally and upload the predictions to obtain the AUC value. \nNow we have put together all our processing pieces and made it a pipeline to submit for the final testing phase. \nWhen we run this pipeline on a subsample of the validation data, it runs successfully. However, when we run it on all the cases, we are facing errors. \n\nThe error we get is \"Notebook Threw Exception\".\nThe description of the error as on the website is \"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\".\n\nCan someone please help us in this regard ?", "votes": null }, { "id": "1545769", "postDate": "10/15/2021 14:40:52", "content": "I was facing the same problem today. I made a pipeline which contains ensemble of around 10 models and it was running fine when saved and run. But in submission I got notebook error. I tried to freed GPU, RAM etc. After every model prediction. But it not helped me. At last I removed 4 models from my final ensemble submission notebook. Then it worked.
", "rawMarkdown": "I was facing the same problem today. I made a pipeline which contains ensemble of around 10 models and it was running fine when saved and run. But in submission I got notebook error. I tried to freed GPU, RAM etc. After every model prediction. But it not helped me. At last I removed 4 models from my final ensemble submission notebook. Then it worked.", "votes": null }, { "id": "1545832", "postDate": "10/15/2021 15:59:22", "content": "I had the same problem today at the wrong time.
\n\"Notebook Threw Exception\"
In my case, I have an ensemble with 20 models.
\nI felt it wasn't a hardware issue as I made the mini-batch size so small that it didn't exceed the GPU memory.
I learned this the hard way: always test your code end2end on kaggle one week before the deadline. i missed two deadlines this year (birdcall and moa). both were silver solutions..
", "rawMarkdown": "I learned this the hard way: always test your code end2end on kaggle one week before the deadline. i missed two deadlines this year (birdcall and moa). both were silver solutions..", "votes": null }, { "id": "1545977", "postDate": "10/15/2021 18:15:44", "content": "Thanks for the reply. I just have one model. I am not doing any ensembling at all.
", "rawMarkdown": "Thanks for the reply. I just have one model. I am not doing any ensembling at all.", "votes": null }, { "id": "1545980", "postDate": "10/15/2021 18:18:13", "content": "Thanks for the reply. In my case, I suspect the algorithm to be running as I can see it take the same time as it took during regular testing.
", "rawMarkdown": "Thanks for the reply. In my case, I suspect the algorithm to be running as I can see it take the same time as it took during regular testing.", "votes": null } ], "comments": [ { "id": 1545769, "author_name": "boltuzamaki", "author_url": "", "post_date": "10/15/2021 14:40:52", "content": "I was facing the same problem today. I made a pipeline which contains ensemble of around 10 models and it was running fine when saved and run. But in submission I got notebook error. I tried to freed GPU, RAM etc. After every model prediction. But it not helped me. At last I removed 4 models from my final ensemble submission notebook. Then it worked.
", "votes": null, "replies": [ { "id": 1545977, "author_name": "teamalpaca", "author_url": "", "post_date": "10/15/2021 18:15:44", "content": "Thanks for the reply. I just have one model. I am not doing any ensembling at all.
", "votes": null, "replies": [] } ] }, { "id": 1545832, "author_name": "hideyukizushi", "author_url": "", "post_date": "10/15/2021 15:59:22", "content": "I had the same problem today at the wrong time.
\n\"Notebook Threw Exception\"
In my case, I have an ensemble with 20 models.
\nI felt it wasn't a hardware issue as I made the mini-batch size so small that it didn't exceed the GPU memory.
Thanks for the reply. In my case, I suspect the algorithm to be running as I can see it take the same time as it took during regular testing.
", "votes": null, "replies": [] } ] }, { "id": 1545970, "author_name": "", "author_url": "", "post_date": "10/15/2021 18:12:26", "content": "I learned this the hard way: always test your code end2end on kaggle one week before the deadline. i missed two deadlines this year (birdcall and moa). both were silver solutions..
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1544855": "Hello,\n\nAll these days, we used to test our networks internally and upload the predictions to obtain the AUC value. \nNow we have put together all our processing pieces and made it a pipeline to submit for the final testing phase. \nWhen we run this pipeline on a subsample of the validation data, it runs successfully. However, when we run it on all the cases, we are facing errors. \n\nThe error we get is \"Notebook Threw Exception\".\nThe description of the error as on the website is \"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\".\n\nCan someone please help us in this regard ?", "1545769": "I was facing the same problem today. I made a pipeline which contains ensemble of around 10 models and it was running fine when saved and run. But in submission I got notebook error. I tried to freed GPU, RAM etc. After every model prediction. But it not helped me. At last I removed 4 models from my final ensemble submission notebook. Then it worked.", "1545832": "I had the same problem today at the wrong time.\n\"Notebook Threw Exception\"\n\nIn my case, I have an ensemble with 20 models.\nI felt it wasn't a hardware issue as I made the mini-batch size so small that it didn't exceed the GPU memory.", "1545970": "I learned this the hard way: always test your code end2end on kaggle one week before the deadline. i missed two deadlines this year (birdcall and moa). both were silver solutions..", "1545977": "Thanks for the reply. I just have one model. I am not doing any ensembling at all.", "1545980": "Thanks for the reply. In my case, I suspect the algorithm to be running as I can see it take the same time as it took during regular testing." }, "source": "meta" }