I know this is a self explained error but I always have this problem no matter what I change. This is the notebook that causes this https://www.kaggle.com/npitsillos/mask-rcnn-w-resnetv2-50-train/notebook?scriptVersionId=81543743. What is odd is that the batch size is 2 images of size 224x224 with the only transform being normalise since I believe albumentations was causing this initially. I have removed it and I still get this though.
\nAny thoughts? Am I missing something straightforward here?
", "messages": [ { "id": 1606003, "postDate": "2021-12-04T19:23:05.573Z", "content": "I know this is a self explained error but I always have this problem no matter what I change. This is the notebook that causes this https://www.kaggle.com/npitsillos/mask-rcnn-w-resnetv2-50-train/notebook?scriptVersionId=81543743. What is odd is that the batch size is 2 images of size 224x224 with the only transform being normalise since I believe albumentations was causing this initially. I have removed it and I still get this though.
\nAny thoughts? Am I missing something straightforward here?
", "rawMarkdown": "I know this is a self explained error but I always have this problem no matter what I change. This is the notebook that causes this https://www.kaggle.com/npitsillos/mask-rcnn-w-resnetv2-50-train/notebook?scriptVersionId=81543743. What is odd is that the batch size is 2 images of size 224x224 with the only transform being normalise since I believe albumentations was causing this initially. I have removed it and I still get this though.\n\nAny thoughts? Am I missing something straightforward here?", "votes": 5 }, { "id": 1610888, "postDate": "2021-12-07T15:24:12.710Z", "content": "", "rawMarkdown": "[solution](https://www.kaggle.com/c/sartorius-cell-instance-segmentation/discussion/293894)", "votes": 1, "replies": [ { "id": 1611027, "postDate": "2021-12-07T17:05:18.983Z", "content": "Thank you!
", "rawMarkdown": "Thank you!" } ] }, { "id": 1606865, "postDate": "2021-12-05T10:44:43.083Z", "content": "Try to add the following in the beginning of the notebook:
\nimport os\nos.environ[\"PYTORCH_NO_CUDA_MEMORY_CACHING\"] = str(1)\nit is not going to help if you need a lot of memory, but if you need just a little bit more then it may help by adding a small room by disabling a cache (it is likely to slow down your notebook as well, unfortunately. Slow is better than \"out of memory\" though).
", "rawMarkdown": "Try to add the following in the beginning of the notebook:\n\n```Python\nimport os\nos.environ[\"PYTORCH_NO_CUDA_MEMORY_CACHING\"] = str(1)\n```\nit is not going to help if you need a lot of memory, but if you need just a little bit more then it may help by adding a small room by disabling a cache (it is likely to slow down your notebook as well, unfortunately. Slow is better than \"out of memory\" though).", "votes": 1, "replies": [ { "id": 1606968, "postDate": "2021-12-05T12:38:46.740Z", "content": "Thank you for your comment! I will try this and see however using a batch size of 2 with image size reduced should do the job. I am starting to think this is the model itself or a problem with pytorch.
\nhttps://github.com/pytorch/pytorch/issues/16417
", "rawMarkdown": "Thank you for your comment! I will try this and see however using a batch size of 2 with image size reduced should do the job. I am starting to think this is the model itself or a problem with pytorch.\n\nhttps://github.com/pytorch/pytorch/issues/16417" } ] }, { "id": 1606966, "postDate": "2021-12-05T12:38:05.253Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 1610888, "author_name": "Hannah B", "author_url": "", "post_date": "2021-12-07T15:24:12.710000", "content": "", "votes": 1, "replies": [ { "id": 1611027, "author_name": "Pitsillos", "author_url": "", "post_date": "2021-12-07T17:05:18.983000", "content": "Thank you!
", "votes": 0, "replies": [] } ] }, { "id": 1606865, "author_name": "Sentinel-1", "author_url": "", "post_date": "2021-12-05T10:44:43.083000", "content": "Try to add the following in the beginning of the notebook:
\nimport os\nos.environ[\"PYTORCH_NO_CUDA_MEMORY_CACHING\"] = str(1)\nit is not going to help if you need a lot of memory, but if you need just a little bit more then it may help by adding a small room by disabling a cache (it is likely to slow down your notebook as well, unfortunately. Slow is better than \"out of memory\" though).
", "votes": 1, "replies": [ { "id": 1606968, "author_name": "Pitsillos", "author_url": "", "post_date": "2021-12-05T12:38:46.740000", "content": "Thank you for your comment! I will try this and see however using a batch size of 2 with image size reduced should do the job. I am starting to think this is the model itself or a problem with pytorch.
\n", "votes": 0, "replies": [] } ] }, { "id": 1606966, "author_name": "", "author_url": "", "post_date": "2021-12-05T12:38:05.253000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1606003": "I know this is a self explained error but I always have this problem no matter what I change. This is the notebook that causes this https://www.kaggle.com/npitsillos/mask-rcnn-w-resnetv2-50-train/notebook?scriptVersionId=81543743. What is odd is that the batch size is 2 images of size 224x224 with the only transform being normalise since I believe albumentations was causing this initially. I have removed it and I still get this though.\n\nAny thoughts? Am I missing something straightforward here?", "1610888": "[solution](https://www.kaggle.com/c/sartorius-cell-instance-segmentation/discussion/293894)", "1606865": "Try to add the following in the beginning of the notebook:\n\n```Python\nimport os\nos.environ[\"PYTORCH_NO_CUDA_MEMORY_CACHING\"] = str(1)\n```\nit is not going to help if you need a lot of memory, but if you need just a little bit more then it may help by adding a small room by disabling a cache (it is likely to slow down your notebook as well, unfortunately. Slow is better than \"out of memory\" though).", "1606966": "" } }