{
  "id": 111514,
  "title": "senet154 throws 'No space left on device' error. Help?",
  "url": "/competitions/understanding_cloud_organization/discussion/111514",
  "author_name": "",
  "post_date": "2019-10-06T13:59:39.595754Z",
  "votes": 2,
  "comment_count": 5,
  "views": 0,
  "content": "<p>I am trying to run senet154 on Kaggle notebook and after 1st epoch, when its trying to save history in tmp file, it throws 'fd 79 failed with No space left on device' error. Can anyone help me here?</p>",
  "messages": [
    {
      "id": "642701",
      "postDate": "10/06/2019 13:59:39",
      "content": "<p>I am trying to run senet154 on Kaggle notebook and after 1st epoch, when its trying to save history in tmp file, it throws 'fd 79 failed with No space left on device' error. Can anyone help me here?</p>",
      "rawMarkdown": "I am trying to run senet154 on Kaggle notebook and after 1st epoch, when its trying to save history in tmp file, it throws 'fd 79 failed with No space left on device' error. Can anyone help me here?",
      "votes": null
    },
    {
      "id": "642702",
      "postDate": "10/06/2019 14:01:22",
      "content": "<p>Sadly kaggle kernels have quite a small disk space...\nWhat framework are you using?</p>",
      "rawMarkdown": "Sadly kaggle kernels have quite a small disk space...\nWhat framework are you using?",
      "votes": null
    },
    {
      "id": "642752",
      "postDate": "10/06/2019 15:26:57",
      "content": "<p>thanks for the reply <a href=\"/artgor\">@artgor</a>, I am using the popular notebook shared by you only :) thanks for that..I am using reduced size images and then trying senet154 from 'qubvel/segmentation_models.pytorch'</p>",
      "rawMarkdown": "thanks for the reply @artgor, I am using the popular notebook shared by you only :) thanks for that..I am using reduced size images and then trying senet154 from 'qubvel/segmentation_models.pytorch'",
      "votes": null
    },
    {
      "id": "642753",
      "postDate": "10/06/2019 15:31:46",
      "content": "<p>Then maybe I have a solution :)</p>\n\n<p>While working on the classification kernel, I also encountered a similar problem. Here is how I solved it:</p>\n\n<p>I wrote <code>CustomCheckpointCallback</code> in my utility script <a href=\"https://www.kaggle.com/artgor/pytorch-utils-for-images\">https://www.kaggle.com/artgor/pytorch-utils-for-images</a> (at the very bottom of the script)</p>\n\n<p>Then I import it and add to the list of Callbacks like this:\n<code>\ncallbacks = [AUCCallback(class_names=['Fish', 'Flower', 'Gravel', 'Sugar'], num_classes=4), EarlyStoppingCallback(patience=5, min_delta=0.001), CriterionCallback(), CustomCheckpointCallback()]\n</code></p>\n\n<p>This callback decreases the disk space required for logging.</p>",
      "rawMarkdown": "Then maybe I have a solution :)\n\nWhile working on the classification kernel, I also encountered a similar problem. Here is how I solved it:\n\nI wrote `CustomCheckpointCallback` in my utility script https://www.kaggle.com/artgor/pytorch-utils-for-images (at the very bottom of the script)\n\nThen I import it and add to the list of Callbacks like this:\n```\ncallbacks = [AUCCallback(class_names=['Fish', 'Flower', 'Gravel', 'Sugar'], num_classes=4), EarlyStoppingCallback(patience=5, min_delta=0.001), CriterionCallback(), CustomCheckpointCallback()]\n```\n\nThis callback decreases the disk space required for logging.",
      "votes": null
    },
    {
      "id": "642807",
      "postDate": "10/06/2019 17:17:03",
      "content": "<p>wow ! this is great <a href=\"/artgor\">@artgor</a> ! thanks a ton ! I'm trying this and will let you know as soon it works.</p>",
      "rawMarkdown": "wow ! this is great @artgor ! thanks a ton ! I'm trying this and will let you know as soon it works.",
      "votes": null
    },
    {
      "id": "642878",
      "postDate": "10/06/2019 18:54:03",
      "content": "<p><a href=\"/artgor\">@artgor</a> it's working. Brilliant solution ! thanks a lot :) </p>",
      "rawMarkdown": "artgor it's working. Brilliant solution ! thanks a lot :)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 642702,
      "author_name": "artgor",
      "author_url": "",
      "post_date": "10/06/2019 14:01:22",
      "content": "<p>Sadly kaggle kernels have quite a small disk space...\nWhat framework are you using?</p>",
      "votes": null,
      "replies": [
        {
          "id": 642752,
          "author_name": "anuragtr",
          "author_url": "",
          "post_date": "10/06/2019 15:26:57",
          "content": "<p>thanks for the reply <a href=\"/artgor\">@artgor</a>, I am using the popular notebook shared by you only :) thanks for that..I am using reduced size images and then trying senet154 from 'qubvel/segmentation_models.pytorch'</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 642753,
          "author_name": "artgor",
          "author_url": "",
          "post_date": "10/06/2019 15:31:46",
          "content": "<p>Then maybe I have a solution :)</p>\n\n<p>While working on the classification kernel, I also encountered a similar problem. Here is how I solved it:</p>\n\n<p>I wrote <code>CustomCheckpointCallback</code> in my utility script <a href=\"https://www.kaggle.com/artgor/pytorch-utils-for-images\">https://www.kaggle.com/artgor/pytorch-utils-for-images</a> (at the very bottom of the script)</p>\n\n<p>Then I import it and add to the list of Callbacks like this:\n<code>\ncallbacks = [AUCCallback(class_names=['Fish', 'Flower', 'Gravel', 'Sugar'], num_classes=4), EarlyStoppingCallback(patience=5, min_delta=0.001), CriterionCallback(), CustomCheckpointCallback()]\n</code></p>\n\n<p>This callback decreases the disk space required for logging.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 642807,
          "author_name": "anuragtr",
          "author_url": "",
          "post_date": "10/06/2019 17:17:03",
          "content": "<p>wow ! this is great <a href=\"/artgor\">@artgor</a> ! thanks a ton ! I'm trying this and will let you know as soon it works.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 642878,
          "author_name": "anuragtr",
          "author_url": "",
          "post_date": "10/06/2019 18:54:03",
          "content": "<p><a href=\"/artgor\">@artgor</a> it's working. Brilliant solution ! thanks a lot :) </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "642701": "I am trying to run senet154 on Kaggle notebook and after 1st epoch, when its trying to save history in tmp file, it throws 'fd 79 failed with No space left on device' error. Can anyone help me here?",
    "642702": "Sadly kaggle kernels have quite a small disk space...\nWhat framework are you using?",
    "642752": "thanks for the reply @artgor, I am using the popular notebook shared by you only :) thanks for that..I am using reduced size images and then trying senet154 from 'qubvel/segmentation_models.pytorch'",
    "642753": "Then maybe I have a solution :)\n\nWhile working on the classification kernel, I also encountered a similar problem. Here is how I solved it:\n\nI wrote `CustomCheckpointCallback` in my utility script https://www.kaggle.com/artgor/pytorch-utils-for-images (at the very bottom of the script)\n\nThen I import it and add to the list of Callbacks like this:\n```\ncallbacks = [AUCCallback(class_names=['Fish', 'Flower', 'Gravel', 'Sugar'], num_classes=4), EarlyStoppingCallback(patience=5, min_delta=0.001), CriterionCallback(), CustomCheckpointCallback()]\n```\n\nThis callback decreases the disk space required for logging.",
    "642807": "wow ! this is great @artgor ! thanks a ton ! I'm trying this and will let you know as soon it works.",
    "642878": "artgor it's working. Brilliant solution ! thanks a lot :)"
  },
  "source": "meta"
}