Hello everyone !\nI am getting this error when I try to submit : \"Notebook Exceeded Allowed Compute\".\nAfter my research I found in this issue that it's may be because my notebook consumes either too much disk or RAM (or GPU Memory I gess). \nSo Does anyone know if their are particular Memory limits in this compétition ? and what are the thresholds ?\nThanks for answering !
", "messages": [ { "id": "876438", "postDate": "06/06/2020 17:24:13", "content": "Hello everyone !\nI am getting this error when I try to submit : \"Notebook Exceeded Allowed Compute\".\nAfter my research I found in this issue that it's may be because my notebook consumes either too much disk or RAM (or GPU Memory I gess). \nSo Does anyone know if their are particular Memory limits in this compétition ? and what are the thresholds ?\nThanks for answering !
", "rawMarkdown": "Hello everyone !\nI am getting this error when I try to submit : \"Notebook Exceeded Allowed Compute\".\nAfter my research I found in [this issue](https://www.kaggle.com/questions-and-answers/129198) that it's may be because my notebook consumes either too much disk or RAM (or GPU Memory I gess). \nSo Does anyone know if their are particular Memory limits in this compétition ? and what are the thresholds ?\nThanks for answering !", "votes": null }, { "id": "876477", "postDate": "06/06/2020 18:00:08", "content": "same thresholds for RAM as for a standard commit\nadditionally:\n\nCPU Notebook <= 9 hours run-time\nGPU Notebook <= 6 hours run-time\n\nas seen here
thanks for your answer,\nIt seems that since we have a hidden test set for the scoring (which is bigger than the test set that comes with the dataset) making prediction directly on the hole set take too much memory (during the scoring). So now I make predictions for images in the set one by one. And It works for me 😁
", "rawMarkdown": "thanks for your answer,\nIt seems that since we have a hidden test set for the scoring (which is bigger than the test set that comes with the dataset) making prediction directly on the hole set take too much memory (during the scoring). So now I make predictions for images in the set one by one. And It works for me 😁", "votes": null }, { "id": "877054", "postDate": "06/07/2020 09:29:40", "content": "correct, predictions should also be done with a DataLoader and batches. We have 1276 images in the test set
", "rawMarkdown": "correct, predictions should also be done with a DataLoader and batches. We have 1276 images in the test set", "votes": null }, { "id": "877278", "postDate": "06/07/2020 13:08:17", "content": "Ok I didn't nkow the exact number, Thanks for that precision
", "rawMarkdown": "Ok I didn't nkow the exact number, Thanks for that precision", "votes": null } ], "comments": [ { "id": 876477, "author_name": "ilu000", "author_url": "", "post_date": "06/06/2020 18:00:08", "content": "same thresholds for RAM as for a standard commit\nadditionally:\n\nCPU Notebook <= 9 hours run-time\nGPU Notebook <= 6 hours run-time\n\nas seen here
thanks for your answer,\nIt seems that since we have a hidden test set for the scoring (which is bigger than the test set that comes with the dataset) making prediction directly on the hole set take too much memory (during the scoring). So now I make predictions for images in the set one by one. And It works for me 😁
", "votes": null, "replies": [] }, { "id": 877054, "author_name": "ilu000", "author_url": "", "post_date": "06/07/2020 09:29:40", "content": "correct, predictions should also be done with a DataLoader and batches. We have 1276 images in the test set
", "votes": null, "replies": [] }, { "id": 877278, "author_name": "layediop", "author_url": "", "post_date": "06/07/2020 13:08:17", "content": "Ok I didn't nkow the exact number, Thanks for that precision
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "876438": "Hello everyone !\nI am getting this error when I try to submit : \"Notebook Exceeded Allowed Compute\".\nAfter my research I found in [this issue](https://www.kaggle.com/questions-and-answers/129198) that it's may be because my notebook consumes either too much disk or RAM (or GPU Memory I gess). \nSo Does anyone know if their are particular Memory limits in this compétition ? and what are the thresholds ?\nThanks for answering !", "876477": "same thresholds for RAM as for a standard commit\nadditionally:\n```\nCPU Notebook <= 9 hours run-time\nGPU Notebook <= 6 hours run-time\n```\nas seen [here](https://www.kaggle.com/c/global-wheat-detection/overview/code-requirements)", "876681": "thanks for your answer,\nIt seems that since we have a hidden test set for the scoring (which is bigger than the test set that comes with the dataset) making prediction directly on the hole set take too much memory (during the scoring). So now I make predictions for images in the set one by one. And It works for me 😁", "877054": "correct, predictions should also be done with a DataLoader and batches. We have 1276 images in the test set", "877278": "Ok I didn't nkow the exact number, Thanks for that precision" }, "source": "meta" }