I have the problem that when I finished my commit and it generated the submission.csv, it cannot calculate the public score. It finally showed the time out. \nIs there someone coming across the same situation?
\nI need help...thx
I have the problem that when I finished my commit and it generated the submission.csv, it cannot calculate the public score. It finally showed the time out. \nIs there someone coming across the same situation?
\nI need help...thx
Your Notebook might have exceeded 2 hours of GPU time.
\n\nTest Images used in LB calculation contains lot many images as provided in actual test set.\nTry optimizing your code or try to do prediction on train set rather than test set to find out areas of improvement.
", "rawMarkdown": "Your Notebook might have exceeded 2 hours of GPU time.\n\nTest Images used in LB calculation contains lot many images as provided in actual test set.\nTry optimizing your code or try to do prediction on train set rather than test set to find out areas of improvement.", "votes": null }, { "id": "713170", "postDate": "01/08/2020 01:42:05", "content": "Hi Amit,\n I also met the question timeout. But when i'm testing, i found that the time of reading parquet is so slow. I upload my trained model of densenet121 to the kernel, so most time is on the reading parquet. Do you know how to solve it?
", "rawMarkdown": "Hi Amit,\n I also met the question timeout. But when i'm testing, i found that the time of reading parquet is so slow. I upload my trained model of densenet121 to the kernel, so most time is on the reading parquet. Do you know how to solve it?", "votes": null }, { "id": "713176", "postDate": "01/08/2020 01:58:10", "content": "Hi ang, we can use feather files for train images but we have to use same test images parquet files for final processing.
\n\nI don't think reading parquet file is taking more than 2 3 mins. Are you doing some another operation like resizing or merging along with reading of parquet file?
", "rawMarkdown": "Hi ang, we can use feather files for train images but we have to use same test images parquet files for final processing. \n\nI don't think reading parquet file is taking more than 2 3 mins. Are you doing some another operation like resizing or merging along with reading of parquet file?", "votes": null }, { "id": "713192", "postDate": "01/08/2020 02:26:09", "content": "Hi Amit,\n I tested the time of reading parquet just now, 90 seconds every parquet. But i do not use the dataset api, just read every image, crop, resize, normalize and send to the model. I think maybe the problem is that i do not use the dataset api, since the api process the dataset in multi-threads. Thank for your help.
", "rawMarkdown": "Hi Amit,\n I tested the time of reading parquet just now, 90 seconds every parquet. But i do not use the dataset api, just read every image, crop, resize, normalize and send to the model. I think maybe the problem is that i do not use the dataset api, since the api process the dataset in multi-threads. Thank for your help.", "votes": null }, { "id": "713742", "postDate": "01/08/2020 15:45:32", "content": "I had that problem too. My problem was I had the GPU off, it was fast when committing because there are only 12 instances to predict. Make sure it is ON
", "rawMarkdown": "I had that problem too. My problem was I had the GPU off, it was fast when committing because there are only 12 instances to predict. Make sure it is ON", "votes": null } ], "comments": [ { "id": 712804, "author_name": "amit9484", "author_url": "", "post_date": "01/07/2020 15:48:49", "content": "Your Notebook might have exceeded 2 hours of GPU time.
\n\nTest Images used in LB calculation contains lot many images as provided in actual test set.\nTry optimizing your code or try to do prediction on train set rather than test set to find out areas of improvement.
", "votes": null, "replies": [ { "id": 713170, "author_name": "liang23333", "author_url": "", "post_date": "01/08/2020 01:42:05", "content": "Hi Amit,\n I also met the question timeout. But when i'm testing, i found that the time of reading parquet is so slow. I upload my trained model of densenet121 to the kernel, so most time is on the reading parquet. Do you know how to solve it?
", "votes": null, "replies": [] }, { "id": 713176, "author_name": "amit9484", "author_url": "", "post_date": "01/08/2020 01:58:10", "content": "Hi ang, we can use feather files for train images but we have to use same test images parquet files for final processing.
\n\nI don't think reading parquet file is taking more than 2 3 mins. Are you doing some another operation like resizing or merging along with reading of parquet file?
", "votes": null, "replies": [] }, { "id": 713192, "author_name": "liang23333", "author_url": "", "post_date": "01/08/2020 02:26:09", "content": "Hi Amit,\n I tested the time of reading parquet just now, 90 seconds every parquet. But i do not use the dataset api, just read every image, crop, resize, normalize and send to the model. I think maybe the problem is that i do not use the dataset api, since the api process the dataset in multi-threads. Thank for your help.
", "votes": null, "replies": [] } ] }, { "id": 713742, "author_name": "macarrony00", "author_url": "", "post_date": "01/08/2020 15:45:32", "content": "I had that problem too. My problem was I had the GPU off, it was fast when committing because there are only 12 instances to predict. Make sure it is ON
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "712729": "I have the problem that when I finished my commit and it generated the submission.csv, it cannot calculate the public score. It finally showed the time out. \nIs there someone coming across the same situation? \nI need help...thx", "712804": "Your Notebook might have exceeded 2 hours of GPU time.\n\nTest Images used in LB calculation contains lot many images as provided in actual test set.\nTry optimizing your code or try to do prediction on train set rather than test set to find out areas of improvement.", "713170": "Hi Amit,\n I also met the question timeout. But when i'm testing, i found that the time of reading parquet is so slow. I upload my trained model of densenet121 to the kernel, so most time is on the reading parquet. Do you know how to solve it?", "713176": "Hi ang, we can use feather files for train images but we have to use same test images parquet files for final processing. \n\nI don't think reading parquet file is taking more than 2 3 mins. Are you doing some another operation like resizing or merging along with reading of parquet file?", "713192": "Hi Amit,\n I tested the time of reading parquet just now, 90 seconds every parquet. But i do not use the dataset api, just read every image, crop, resize, normalize and send to the model. I think maybe the problem is that i do not use the dataset api, since the api process the dataset in multi-threads. Thank for your help.", "713742": "I had that problem too. My problem was I had the GPU off, it was fast when committing because there are only 12 instances to predict. Make sure it is ON" }, "source": "meta" }