{
  "id": 191657,
  "title": "MONAI transforms and TPU?",
  "url": "/competitions/rsna-str-pulmonary-embolism-detection/discussion/191657",
  "author_name": "Araik Tamazian",
  "post_date": "2020-10-17T18:15:57.833000",
  "votes": 0,
  "comment_count": 6,
  "views": 0,
  "content": "<p>I wonder if anyone here was able to utilize MONAI transforms together with PytTorch XLA (TPU). I'm getting <code>ValueError: recompute_scale_factor is not meaningful with an explicit size.</code></p>",
  "messages": [
    {
      "id": 1052711,
      "postDate": "2020-10-18T07:09:09.923Z",
      "content": "<p>I am training it on TPU without any issue </p>\n<p>You need some serious optimization if you use colab though </p>",
      "rawMarkdown": "I am training it on TPU without any issue \n\nYou need some serious optimization if you use colab though ",
      "votes": 1,
      "replies": [
        {
          "id": 1052845,
          "postDate": "2020-10-18T10:50:20.233Z",
          "content": "<p>Do you use Pytorch XLA?</p>",
          "rawMarkdown": "Do you use Pytorch XLA?"
        },
        {
          "id": 1052854,
          "postDate": "2020-10-18T11:00:26.980Z",
          "content": "<p>Yes , but a relatively old version</p>\n<p>I didn't try the last releases. </p>",
          "rawMarkdown": "Yes , but a relatively old version\n\nI didn't try the last releases. ",
          "votes": 1
        },
        {
          "id": 1052881,
          "postDate": "2020-10-18T11:35:56.403Z",
          "content": "<p>I see, thanks for the info.</p>",
          "rawMarkdown": "I see, thanks for the info."
        }
      ]
    },
    {
      "id": 1053042,
      "postDate": "2020-10-18T14:40:22.027Z",
      "content": "<p>I've managed to by solve this issue by changing <code>--version nightly</code> to <code>--version 1.6</code> in the PyTorch XLA installation code.</p>",
      "rawMarkdown": "I've managed to by solve this issue by changing `--version nightly` to `--version 1.6` in the PyTorch XLA installation code.",
      "votes": 2,
      "replies": [
        {
          "id": 1053219,
          "postDate": "2020-10-18T17:48:47.777Z",
          "content": "<p>I got it now! Thanks</p>",
          "rawMarkdown": "I got it now! Thanks"
        }
      ]
    },
    {
      "id": 1052442,
      "postDate": "2020-10-17T18:15:57.833Z",
      "content": "<p>I wonder if anyone here was able to utilize MONAI transforms together with PytTorch XLA (TPU). I'm getting <code>ValueError: recompute_scale_factor is not meaningful with an explicit size.</code></p>",
      "rawMarkdown": "I wonder if anyone here was able to utilize MONAI transforms together with PytTorch XLA (TPU). I'm getting `ValueError: recompute_scale_factor is not meaningful with an explicit size.`"
    }
  ],
  "comments": [
    {
      "id": 1052711,
      "author_name": "Serigne ",
      "author_url": "",
      "post_date": "2020-10-18T07:09:09.923000",
      "content": "<p>I am training it on TPU without any issue </p>\n<p>You need some serious optimization if you use colab though </p>",
      "votes": 1,
      "replies": [
        {
          "id": 1052845,
          "author_name": "Araik Tamazian",
          "author_url": "",
          "post_date": "2020-10-18T10:50:20.233000",
          "content": "<p>Do you use Pytorch XLA?</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1052854,
          "author_name": "Serigne ",
          "author_url": "",
          "post_date": "2020-10-18T11:00:26.980000",
          "content": "<p>Yes , but a relatively old version</p>\n<p>I didn't try the last releases. </p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 1052881,
          "author_name": "Araik Tamazian",
          "author_url": "",
          "post_date": "2020-10-18T11:35:56.403000",
          "content": "<p>I see, thanks for the info.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1053042,
      "author_name": "Araik Tamazian",
      "author_url": "",
      "post_date": "2020-10-18T14:40:22.027000",
      "content": "<p>I've managed to by solve this issue by changing <code>--version nightly</code> to <code>--version 1.6</code> in the PyTorch XLA installation code.</p>",
      "votes": 2,
      "replies": [
        {
          "id": 1053219,
          "author_name": "Ronaldo S.A. Batista",
          "author_url": "",
          "post_date": "2020-10-18T17:48:47.777000",
          "content": "<p>I got it now! Thanks</p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1052711": "I am training it on TPU without any issue \n\nYou need some serious optimization if you use colab though ",
    "1053042": "I've managed to by solve this issue by changing `--version nightly` to `--version 1.6` in the PyTorch XLA installation code.",
    "1052442": "I wonder if anyone here was able to utilize MONAI transforms together with PytTorch XLA (TPU). I'm getting `ValueError: recompute_scale_factor is not meaningful with an explicit size.`"
  }
}