![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F5032548%2F15e83a49ee93498b1ab2076c3d4e50ba%2FAnnotation%202020-06-16%20022559.png?generation=1592254607413162&alt=media\")
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F5032548%2F7eba8a7daab53cdabce1fd41a6a80e94%2FAnnotation%202020-06-16%20022447.png?generation=1592254614317354&alt=media\")
Using tensorflow 2.2.0 with GPU as accelerator, but only CPU is getting used.\nAttaching images.
\n\n\n
So the only solution is to use resized images?
", "rawMarkdown": "So the only solution is to use resized images?", "replies": [ { "id": 889026, "postDate": "2020-06-16T18:19:11.220Z", "content": "Resizing your images will only make image processing longer, although it will make gpu learning quicker. I recommended a smaller batch size so you wouldn't have to wait long to see that the gpu is actually working. The GPU is only used for fitting, not loading the images. The cpu loads the images which is why its usage is so high.
", "rawMarkdown": "Resizing your images will only make image processing longer, although it will make gpu learning quicker. I recommended a smaller batch size so you wouldn't have to wait long to see that the gpu is actually working. The GPU is only used for fitting, not loading the images. The cpu loads the images which is why its usage is so high." }, { "id": 889093, "postDate": "2020-06-16T18:53:48.180Z", "content": "It was talking about using an already resized dataset. I think 224X224 is available.\nAlso I am currently using a batch size of only 20, I can reduce it a bit further, if that would help.
", "rawMarkdown": "It was talking about using an already resized dataset. I think 224X224 is available.\nAlso I am currently using a batch size of only 20, I can reduce it a bit further, if that would help." }, { "id": 889325, "postDate": "2020-06-16T23:03:40.247Z", "content": "If you have the ability to load images quicker with little accuracy trade-off, then a resized dataset sounds like a good course of action.
", "rawMarkdown": "If you have the ability to load images quicker with little accuracy trade-off, then a resized dataset sounds like a good course of action." } ] }, { "id": 887885, "postDate": "2020-06-16T01:11:13.190Z", "content": "Sometimes image loading takes a long time, put a smaller batch size and watch the gpu usage, you'll probably see it go up for a couple seconds and then see it return to 0% again.
", "rawMarkdown": "Sometimes image loading takes a long time, put a smaller batch size and watch the gpu usage, you'll probably see it go up for a couple seconds and then see it return to 0% again." }, { "id": 887755, "postDate": "2020-06-15T20:57:20.130Z", "content": "Using tensorflow 2.2.0 with GPU as accelerator, but only CPU is getting used.\nAttaching images.
\n\n\n
So the only solution is to use resized images?
", "votes": 0, "replies": [ { "id": 889026, "author_name": "Pythonian", "author_url": "", "post_date": "2020-06-16T18:19:11.220000", "content": "Resizing your images will only make image processing longer, although it will make gpu learning quicker. I recommended a smaller batch size so you wouldn't have to wait long to see that the gpu is actually working. The GPU is only used for fitting, not loading the images. The cpu loads the images which is why its usage is so high.
", "votes": 0, "replies": [] }, { "id": 889093, "author_name": "Satyam Tripathi 123", "author_url": "", "post_date": "2020-06-16T18:53:48.180000", "content": "It was talking about using an already resized dataset. I think 224X224 is available.\nAlso I am currently using a batch size of only 20, I can reduce it a bit further, if that would help.
", "votes": 0, "replies": [] }, { "id": 889325, "author_name": "Pythonian", "author_url": "", "post_date": "2020-06-16T23:03:40.247000", "content": "If you have the ability to load images quicker with little accuracy trade-off, then a resized dataset sounds like a good course of action.
", "votes": 0, "replies": [] } ] }, { "id": 887885, "author_name": "Pythonian", "author_url": "", "post_date": "2020-06-16T01:11:13.190000", "content": "Sometimes image loading takes a long time, put a smaller batch size and watch the gpu usage, you'll probably see it go up for a couple seconds and then see it return to 0% again.
", "votes": 0, "replies": [] }, { "id": 888316, "author_name": "", "author_url": "", "post_date": "2020-06-16T09:14:48.587000", "content": "", "votes": 0, "replies": [] }, { "id": 887884, "author_name": "", "author_url": "", "post_date": "2020-06-16T01:10:44.263000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "888121": "So the only solution is to use resized images?", "887885": "Sometimes image loading takes a long time, put a smaller batch size and watch the gpu usage, you'll probably see it go up for a couple seconds and then see it return to 0% again.", "887755": "Using tensorflow 2.2.0 with GPU as accelerator, but only CPU is getting used.\nAttaching images.\n\n\n\n", "888316": "", "887884": "" } }