{
  "id": 550323,
  "title": "Problem for model",
  "url": "/competitions/czii-cryo-et-object-identification/discussion/550323",
  "author_name": "",
  "post_date": "2024-12-06T18:50:45.151158500Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I created the masks like this:<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F5591682%2F67dbcb818d58dae9e9a1aa6710a5382f%2Foutput.png?generation=1733510974242522&amp;alt=media\" alt=\"\"></p>\n<p>Is it correct?</p>\n<p>When running my model I encountered the problem \"… GPU out of allocated memory\". What should I do?</p>",
  "messages": [
    {
      "id": "3065397",
      "postDate": "12/06/2024 18:50:45",
      "content": "<p>I created the masks like this:<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F5591682%2F67dbcb818d58dae9e9a1aa6710a5382f%2Foutput.png?generation=1733510974242522&amp;alt=media\" alt=\"\"></p>\n<p>Is it correct?</p>\n<p>When running my model I encountered the problem \"… GPU out of allocated memory\". What should I do?</p>",
      "rawMarkdown": "I created the masks like this:![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F5591682%2F67dbcb818d58dae9e9a1aa6710a5382f%2Foutput.png?generation=1733510974242522&alt=media)\n\nIs it correct?\n\nWhen running my model I encountered the problem \"... GPU out of allocated memory\". What should I do?",
      "votes": null
    },
    {
      "id": "3065466",
      "postDate": "12/06/2024 20:38:16",
      "content": "<p>Reduce memory usage.</p>\n<p>EDIT: The batch size, the dimensions of the patches on each batch, the model parameters,ensure model and data in float16, ensure fold in with torch.no_grad() for inference only; I almost forgot, try GPU T4 x2,… The mask looks correct.</p>",
      "rawMarkdown": "Reduce memory usage.\n\nEDIT: The batch size, the dimensions of the patches on each batch, the model parameters,ensure model and data in float16, ensure fold in with torch.no_grad() for inference only; I almost forgot, try GPU T4 x2,... The mask looks correct.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3065466,
      "author_name": "sacuscreed",
      "author_url": "",
      "post_date": "12/06/2024 20:38:16",
      "content": "<p>Reduce memory usage.</p>\n<p>EDIT: The batch size, the dimensions of the patches on each batch, the model parameters,ensure model and data in float16, ensure fold in with torch.no_grad() for inference only; I almost forgot, try GPU T4 x2,… The mask looks correct.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3065397": "I created the masks like this:![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F5591682%2F67dbcb818d58dae9e9a1aa6710a5382f%2Foutput.png?generation=1733510974242522&alt=media)\n\nIs it correct?\n\nWhen running my model I encountered the problem \"... GPU out of allocated memory\". What should I do?",
    "3065466": "Reduce memory usage.\n\nEDIT: The batch size, the dimensions of the patches on each batch, the model parameters,ensure model and data in float16, ensure fold in with torch.no_grad() for inference only; I almost forgot, try GPU T4 x2,... The mask looks correct."
  },
  "source": "meta"
}