{ "id": 406210, "title": "Notebook timeout issue", "url": "/competitions/vesuvius-challenge-ink-detection/discussion/406210", "author_name": "", "post_date": "2023-05-01T12:30:22.011265100Z", "votes": 1, "comment_count": 4, "views": 0, "content": "

Hello, my prediction model spent around 5 hours to predict all pixels in the test set (the trained model is loaded directly thus no training time).

\n

The saved notebook was success and I can even download submission.csv for fragment a and b. However, it appears as Notebook Timeout in submission results ๐Ÿ˜ญ.

\n

So my questions are

\n
    \n
  1. What is the time limit for submissions? I tried to find the time limit in Rules, but can't find any.
    2. \n
  2. If the problem is really time limitation, can you give me the tips to optimize the model ๐Ÿฅน.
    3. \n
\n

Thanks in advance guys!

", "messages": [ { "id": "2241358", "postDate": "05/01/2023 12:30:22", "content": "

Hello, my prediction model spent around 5 hours to predict all pixels in the test set (the trained model is loaded directly thus no training time).

\n

The saved notebook was success and I can even download submission.csv for fragment a and b. However, it appears as Notebook Timeout in submission results ๐Ÿ˜ญ.

\n

So my questions are

\n
    \n
  1. What is the time limit for submissions? I tried to find the time limit in Rules, but can't find any.
    2. \n
  2. If the problem is really time limitation, can you give me the tips to optimize the model ๐Ÿฅน.
    3. \n
\n

Thanks in advance guys!

", "rawMarkdown": "Hello, my prediction model spent around 5 hours to predict all pixels in the test set (the trained model is loaded directly thus no training time). \n\nThe saved notebook was success and I can even download submission.csv for **fragment a and b**. However, it appears as *Notebook Timeout* in submission results ๐Ÿ˜ญ.\n\nSo my questions are\n1. What is the time limit for submissions? I tried to find the time limit in *Rules*, but can't find any.\n2. If the problem is really time limitation, can you give me the tips to optimize the model ๐Ÿฅน. \n\nThanks in advance guys!", "votes": null }, { "id": "2241445", "postDate": "05/01/2023 14:09:27", "content": "

There is the explanation in there.
\n\"\"

", "rawMarkdown": "**There is the explanation in there.**\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8293010%2Fb6538b95eeb76502fe5369f4567d195e%2F2023-05-01%2022-06-36.png?generation=1682950027988540&alt=media)", "votes": null }, { "id": "2241482", "postDate": "05/01/2023 14:38:05", "content": "

The real hidden test set (not a dummy part: a and b sample) has a much larger resolution.

\n

From the data description page:
\nYou can expect two fragments in the hidden test set, which together are roughly the same size as a single training fragment. The sample slices available to download in the test folders are simply copied from training fragment one, but when you submit your notebook they will be substituted with the real test data.

", "rawMarkdown": "The real hidden test set (not a dummy part: a and b sample) has a much larger resolution. \n\nFrom the data description [page](https://www.kaggle.com/competitions/vesuvius-challenge-ink-detection/data):\n` You can expect two fragments in the hidden test set, which together are roughly the same size as a single training fragment. The sample slices available to download in the test folders are simply copied from training fragment one, but when you submit your notebook they will be substituted with the real test data.`", "votes": null }, { "id": "2241597", "postDate": "05/01/2023 16:24:58", "content": "

Oops thanks! I literally go through rules section as mentioned in the failing page ๐Ÿ˜….

\n

But hmm even 9 hours still doesnโ€™t enough for hidden test set on free GPU ๐Ÿ˜ฑ.

", "rawMarkdown": "Oops thanks! I literally go through rules section as mentioned in the failing page ๐Ÿ˜…. \n\nBut hmm even 9 hours still doesnโ€™t enough for hidden test set on free GPU ๐Ÿ˜ฑ.", "votes": null }, { "id": "2241599", "postDate": "05/01/2023 16:26:59", "content": "

Thanks! Definitely need to find an optimization ๐Ÿ˜…

", "rawMarkdown": "Thanks! Definitely need to find an optimization ๐Ÿ˜…", "votes": null } ], "comments": [ { "id": 2241445, "author_name": "yoyobar", "author_url": "", "post_date": "05/01/2023 14:09:27", "content": "

There is the explanation in there.
\n\"\"

", "votes": null, "replies": [ { "id": 2241597, "author_name": "korntewinboonchuay", "author_url": "", "post_date": "05/01/2023 16:24:58", "content": "

Oops thanks! I literally go through rules section as mentioned in the failing page ๐Ÿ˜….

\n

But hmm even 9 hours still doesnโ€™t enough for hidden test set on free GPU ๐Ÿ˜ฑ.

", "votes": null, "replies": [] } ] }, { "id": 2241482, "author_name": "igorkrashenyi", "author_url": "", "post_date": "05/01/2023 14:38:05", "content": "

The real hidden test set (not a dummy part: a and b sample) has a much larger resolution.

\n

From the data description page:
\nYou can expect two fragments in the hidden test set, which together are roughly the same size as a single training fragment. The sample slices available to download in the test folders are simply copied from training fragment one, but when you submit your notebook they will be substituted with the real test data.

", "votes": null, "replies": [ { "id": 2241599, "author_name": "korntewinboonchuay", "author_url": "", "post_date": "05/01/2023 16:26:59", "content": "

Thanks! Definitely need to find an optimization ๐Ÿ˜…

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "2241358": "Hello, my prediction model spent around 5 hours to predict all pixels in the test set (the trained model is loaded directly thus no training time). \n\nThe saved notebook was success and I can even download submission.csv for **fragment a and b**. However, it appears as *Notebook Timeout* in submission results ๐Ÿ˜ญ.\n\nSo my questions are\n1. What is the time limit for submissions? I tried to find the time limit in *Rules*, but can't find any.\n2. If the problem is really time limitation, can you give me the tips to optimize the model ๐Ÿฅน. \n\nThanks in advance guys!", "2241445": "**There is the explanation in there.**\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F8293010%2Fb6538b95eeb76502fe5369f4567d195e%2F2023-05-01%2022-06-36.png?generation=1682950027988540&alt=media)", "2241482": "The real hidden test set (not a dummy part: a and b sample) has a much larger resolution. \n\nFrom the data description [page](https://www.kaggle.com/competitions/vesuvius-challenge-ink-detection/data):\n` You can expect two fragments in the hidden test set, which together are roughly the same size as a single training fragment. The sample slices available to download in the test folders are simply copied from training fragment one, but when you submit your notebook they will be substituted with the real test data.`", "2241597": "Oops thanks! I literally go through rules section as mentioned in the failing page ๐Ÿ˜…. \n\nBut hmm even 9 hours still doesnโ€™t enough for hidden test set on free GPU ๐Ÿ˜ฑ.", "2241599": "Thanks! Definitely need to find an optimization ๐Ÿ˜…" }, "source": "meta" }