{ "id": 225598, "title": "The inference notebook running from the past 14 hours and it still running ", "url": "/competitions/ranzcr-clip-catheter-line-classification/discussion/225598", "author_name": "", "post_date": "2021-03-13T04:40:01.227473600Z", "votes": 3, "comment_count": 5, "views": 0, "content": "

I submitted a notebook with only one model for inference. It already took 14 hours and still show me running.
\nTo be sure I submitted the same notebook again it gives results in just 15 mins.
\nCan anyone tell me what is going on?

", "messages": [ { "id": "1236381", "postDate": "03/13/2021 04:40:01", "content": "

I submitted a notebook with only one model for inference. It already took 14 hours and still show me running.
\nTo be sure I submitted the same notebook again it gives results in just 15 mins.
\nCan anyone tell me what is going on?

", "rawMarkdown": "I submitted a notebook with only one model for inference. It already took 14 hours and still show me running.\nTo be sure I submitted the same notebook again it gives results in just 15 mins.\nCan anyone tell me what is going on?", "votes": null }, { "id": "1236390", "postDate": "03/13/2021 05:10:31", "content": "

you'll get error on whatever is going on. Still things you should check

\n
    \n
  1. batch size: not too small, not too big.
    2. \n
  2. your notebook is memory efficient, for too many models ensemble you might get error 'notebook exeeded out of memory', best way is to name all models a single particular name(eg predict with one model named 'model', save preds in list, make new model name it 'model' too and so on), this way you save memory by overwriting.
    3. \n
", "rawMarkdown": "you'll get error on whatever is going on. Still things you should check \n1. batch size: not too small, not too big.\n2. your notebook is memory efficient, for too many models ensemble you might get error 'notebook exeeded out of memory', best way is to name all models a single particular name(eg predict with one model named 'model', save preds in list, make new model name it 'model' too and so on), this way you save memory by overwriting.", "votes": null }, { "id": "1236393", "postDate": "03/13/2021 05:21:22", "content": "

Check out this and this discussions. There was a submission problem, now it's resolved.

", "rawMarkdown": "Check out [this](https://www.kaggle.com/c/ranzcr-clip-catheter-line-classification/discussion/225383) and [this](https://www.kaggle.com/c/ranzcr-clip-catheter-line-classification/discussion/225528) discussions. There was a submission problem, now it's resolved.", "votes": null }, { "id": "1236609", "postDate": "03/13/2021 09:48:44", "content": "

Thank you for your reply @anku5hk. But batchaize=64 and only one model is there no resembling.
\nThe same notebook without any changes run 15 min. I just have a question that generally if inference time exceeds then it must have to give me an error but it still shows me running.

", "rawMarkdown": "Thank you for your reply @anku5hk. But batchaize=64 and only one model is there no resembling.\nThe same notebook without any changes run 15 min. I just have a question that generally if inference time exceeds then it must have to give me an error but it still shows me running.", "votes": null }, { "id": "1236615", "postDate": "03/13/2021 09:58:12", "content": "

Thank you @andy1010. It helps a lot.

", "rawMarkdown": "Thank you @andy1010. It helps a lot.", "votes": null }, { "id": "1236840", "postDate": "03/13/2021 13:54:47", "content": "

Okay. As discussed here, the problem seems to be on server side. Maybe they have fixed it already, i also ran a notebook today and it ran fine, so you should re-run it.

", "rawMarkdown": "Okay. As discussed [here](https://www.kaggle.com/c/ranzcr-clip-catheter-line-classification/discussion/225528), the problem seems to be on server side. Maybe they have fixed it already, i also ran a notebook today and it ran fine, so you should re-run it.", "votes": null } ], "comments": [ { "id": 1236390, "author_name": "anku5hk", "author_url": "", "post_date": "03/13/2021 05:10:31", "content": "

you'll get error on whatever is going on. Still things you should check

\n
    \n
  1. batch size: not too small, not too big.
    2. \n
  2. your notebook is memory efficient, for too many models ensemble you might get error 'notebook exeeded out of memory', best way is to name all models a single particular name(eg predict with one model named 'model', save preds in list, make new model name it 'model' too and so on), this way you save memory by overwriting.
    3. \n
", "votes": null, "replies": [ { "id": 1236609, "author_name": "vatsalmavani", "author_url": "", "post_date": "03/13/2021 09:48:44", "content": "

Thank you for your reply @anku5hk. But batchaize=64 and only one model is there no resembling.
\nThe same notebook without any changes run 15 min. I just have a question that generally if inference time exceeds then it must have to give me an error but it still shows me running.

", "votes": null, "replies": [] }, { "id": 1236840, "author_name": "anku5hk", "author_url": "", "post_date": "03/13/2021 13:54:47", "content": "

Okay. As discussed here, the problem seems to be on server side. Maybe they have fixed it already, i also ran a notebook today and it ran fine, so you should re-run it.

", "votes": null, "replies": [] } ] }, { "id": 1236393, "author_name": "andy1010", "author_url": "", "post_date": "03/13/2021 05:21:22", "content": "

Check out this and this discussions. There was a submission problem, now it's resolved.

", "votes": null, "replies": [ { "id": 1236615, "author_name": "vatsalmavani", "author_url": "", "post_date": "03/13/2021 09:58:12", "content": "

Thank you @andy1010. It helps a lot.

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1236381": "I submitted a notebook with only one model for inference. It already took 14 hours and still show me running.\nTo be sure I submitted the same notebook again it gives results in just 15 mins.\nCan anyone tell me what is going on?", "1236390": "you'll get error on whatever is going on. Still things you should check \n1. batch size: not too small, not too big.\n2. your notebook is memory efficient, for too many models ensemble you might get error 'notebook exeeded out of memory', best way is to name all models a single particular name(eg predict with one model named 'model', save preds in list, make new model name it 'model' too and so on), this way you save memory by overwriting.", "1236393": "Check out [this](https://www.kaggle.com/c/ranzcr-clip-catheter-line-classification/discussion/225383) and [this](https://www.kaggle.com/c/ranzcr-clip-catheter-line-classification/discussion/225528) discussions. There was a submission problem, now it's resolved.", "1236609": "Thank you for your reply @anku5hk. But batchaize=64 and only one model is there no resembling.\nThe same notebook without any changes run 15 min. I just have a question that generally if inference time exceeds then it must have to give me an error but it still shows me running.", "1236615": "Thank you @andy1010. It helps a lot.", "1236840": "Okay. As discussed [here](https://www.kaggle.com/c/ranzcr-clip-catheter-line-classification/discussion/225528), the problem seems to be on server side. Maybe they have fixed it already, i also ran a notebook today and it ran fine, so you should re-run it." }, "source": "meta" }