I have had a notebook timeout (+9hrs) even though Running the kernel all the way to the generation of the submission.csv file tool under 1hr. Could the private test set be enormous compared to the public one? Any idea why this might happen?
", "messages": [ { "id": "1445352", "postDate": "08/04/2021 08:36:02", "content": "I have had a notebook timeout (+9hrs) even though Running the kernel all the way to the generation of the submission.csv file tool under 1hr. Could the private test set be enormous compared to the public one? Any idea why this might happen?
", "rawMarkdown": "I have had a notebook timeout (+9hrs) even though Running the kernel all the way to the generation of the submission.csv file tool under 1hr. Could the private test set be enormous compared to the public one? Any idea why this might happen?", "votes": null }, { "id": "1445629", "postDate": "08/04/2021 09:34:23", "content": "I'm not sure about your notebook timeout, but the hidden test set is approximately of the same scale as the train dataset. The competition hosts have mentioned that here
", "rawMarkdown": "I'm not sure about your notebook timeout, but the hidden test set is approximately of the same scale as the train dataset. The competition hosts have mentioned that [here](https://www.kaggle.com/c/siim-covid19-detection/data)", "votes": null }, { "id": "1445674", "postDate": "08/04/2021 09:43:10", "content": "Thanks for your reply. Very strange, running the full code on the kernel up to writing the submission file took ~ 1hr, but the submission/scoring timed out after 9hrs+
", "rawMarkdown": "Thanks for your reply. Very strange, running the full code on the kernel up to writing the submission file took ~ 1hr, but the submission/scoring timed out after 9hrs+", "votes": null }, { "id": "1448239", "postDate": "08/04/2021 17:01:21", "content": "Hi, make sure you use GPU accelerator.
", "rawMarkdown": "Hi, make sure you use GPU accelerator.", "votes": null }, { "id": "1449776", "postDate": "08/05/2021 00:40:51", "content": "GPU accelerator?
", "rawMarkdown": "GPU accelerator?", "votes": null }, { "id": "1450161", "postDate": "08/05/2021 03:54:19", "content": "It happened to me just once in this competition. I just reran the same notebook and then it submitted in 2 hours. So you can try again.
", "rawMarkdown": "It happened to me just once in this competition. I just reran the same notebook and then it submitted in 2 hours. So you can try again.", "votes": null }, { "id": "1455995", "postDate": "08/06/2021 18:54:36", "content": "Yes, of course. The notebook ran in ~2h on the kernel (with GPU), that's why I found it so strange to take 9h then timeout in scoring mode with the private test set that should be comparable in size with the public test set
", "rawMarkdown": "Yes, of course. The notebook ran in ~2h on the kernel (with GPU), that's why I found it so strange to take 9h then timeout in scoring mode with the private test set that should be comparable in size with the public test set", "votes": null }, { "id": "1462357", "postDate": "08/09/2021 19:54:21", "content": "I don’t know what’s wrong, but I could avoid timeout errors with smaller batch size on inference. (That would be an accidental success though.)
", "rawMarkdown": "I don’t know what’s wrong, but I could avoid timeout errors with smaller batch size on inference. (That would be an accidental success though.)", "votes": null }, { "id": "1464253", "postDate": "08/10/2021 14:00:32", "content": "Update: I experienced a timeout with smaller batch size. 😇
", "rawMarkdown": "Update: I experienced a timeout with smaller batch size. 😇", "votes": null } ], "comments": [ { "id": 1445629, "author_name": "rai555", "author_url": "", "post_date": "08/04/2021 09:34:23", "content": "I'm not sure about your notebook timeout, but the hidden test set is approximately of the same scale as the train dataset. The competition hosts have mentioned that here
", "votes": null, "replies": [ { "id": 1445674, "author_name": "constantinciambur", "author_url": "", "post_date": "08/04/2021 09:43:10", "content": "Thanks for your reply. Very strange, running the full code on the kernel up to writing the submission file took ~ 1hr, but the submission/scoring timed out after 9hrs+
", "votes": null, "replies": [] } ] }, { "id": 1448239, "author_name": "yus002", "author_url": "", "post_date": "08/04/2021 17:01:21", "content": "Hi, make sure you use GPU accelerator.
", "votes": null, "replies": [] }, { "id": 1449776, "author_name": "parisadst", "author_url": "", "post_date": "08/05/2021 00:40:51", "content": "GPU accelerator?
", "votes": null, "replies": [ { "id": 1455995, "author_name": "constantinciambur", "author_url": "", "post_date": "08/06/2021 18:54:36", "content": "Yes, of course. The notebook ran in ~2h on the kernel (with GPU), that's why I found it so strange to take 9h then timeout in scoring mode with the private test set that should be comparable in size with the public test set
", "votes": null, "replies": [] } ] }, { "id": 1450161, "author_name": "shashwatraman", "author_url": "", "post_date": "08/05/2021 03:54:19", "content": "It happened to me just once in this competition. I just reran the same notebook and then it submitted in 2 hours. So you can try again.
", "votes": null, "replies": [] }, { "id": 1462357, "author_name": "akihirok", "author_url": "", "post_date": "08/09/2021 19:54:21", "content": "I don’t know what’s wrong, but I could avoid timeout errors with smaller batch size on inference. (That would be an accidental success though.)
", "votes": null, "replies": [ { "id": 1464253, "author_name": "akihirok", "author_url": "", "post_date": "08/10/2021 14:00:32", "content": "Update: I experienced a timeout with smaller batch size. 😇
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1445352": "I have had a notebook timeout (+9hrs) even though Running the kernel all the way to the generation of the submission.csv file tool under 1hr. Could the private test set be enormous compared to the public one? Any idea why this might happen?", "1445629": "I'm not sure about your notebook timeout, but the hidden test set is approximately of the same scale as the train dataset. The competition hosts have mentioned that [here](https://www.kaggle.com/c/siim-covid19-detection/data)", "1445674": "Thanks for your reply. Very strange, running the full code on the kernel up to writing the submission file took ~ 1hr, but the submission/scoring timed out after 9hrs+", "1448239": "Hi, make sure you use GPU accelerator.", "1449776": "GPU accelerator?", "1450161": "It happened to me just once in this competition. I just reran the same notebook and then it submitted in 2 hours. So you can try again.", "1455995": "Yes, of course. The notebook ran in ~2h on the kernel (with GPU), that's why I found it so strange to take 9h then timeout in scoring mode with the private test set that should be comparable in size with the public test set", "1462357": "I don’t know what’s wrong, but I could avoid timeout errors with smaller batch size on inference. (That would be an accidental success though.)", "1464253": "Update: I experienced a timeout with smaller batch size. 😇" }, "source": "meta" }