{
  "id": 152524,
  "title": "Help ! UnavailableError: Socket closed",
  "url": "/competitions/alaska2-image-steganalysis/discussion/152524",
  "author_name": "",
  "post_date": "2020-05-20T08:33:25.509969200Z",
  "votes": null,
  "comment_count": 6,
  "views": 0,
  "content": "<p>when using tpu training, at the end of the first epoch, met this error, any idea?\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F4044110%2F292c5af5778cbc3af8828972b49b25e1%2Ftpu_error.png?generation=1589963588974029&amp;alt=media\" alt=\"\"></p>",
  "messages": [
    {
      "id": "854694",
      "postDate": "05/20/2020 08:33:25",
      "content": "<p>when using tpu training, at the end of the first epoch, met this error, any idea?\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F4044110%2F292c5af5778cbc3af8828972b49b25e1%2Ftpu_error.png?generation=1589963588974029&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "when using tpu training, at the end of the first epoch, met this error, any idea?\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F4044110%2F292c5af5778cbc3af8828972b49b25e1%2Ftpu_error.png?generation=1589963588974029&amp;alt=media)",
      "votes": null
    },
    {
      "id": "855493",
      "postDate": "05/21/2020 01:43:21",
      "content": "<p>Solved by reducing batch size.</p>",
      "rawMarkdown": "Solved by reducing batch size.",
      "votes": null
    },
    {
      "id": "905445",
      "postDate": "06/28/2020 14:56:22",
      "content": "<p><a href=\"/zhenglv\">@zhenglv</a> What batch size did you set at the beginning, and what did you use to solve the problem? Thanks!</p>",
      "rawMarkdown": "zhenglv What batch size did you set at the beginning, and what did you use to solve the problem? Thanks!",
      "votes": null
    },
    {
      "id": "905944",
      "postDate": "06/29/2020 01:21:16",
      "content": "<p>64 * 8  failed , 32 * 8  succeeded</p>",
      "rawMarkdown": "64 * 8  failed , 32 * 8  succeeded",
      "votes": null
    },
    {
      "id": "905957",
      "postDate": "06/29/2020 01:55:29",
      "content": "<p><a href=\"/zhenglv\">@zhenglv</a> That's weird. I even have the error when having 8 * 8 in colab. solved the error by removing .cache(), but training went very slow that way.</p>",
      "rawMarkdown": "zhenglv That's weird. I even have the error when having 8 * 8 in colab. solved the error by removing .cache(), but training went very slow that way.",
      "votes": null
    },
    {
      "id": "907450",
      "postDate": "06/30/2020 01:11:38",
      "content": "<p>Have you changed the pics into tfrecords ? After changing them into tfrecords , my note runs a epoch about 10 mins.</p>",
      "rawMarkdown": "Have you changed the pics into tfrecords ? After changing them into tfrecords , my note runs a epoch about 10 mins.",
      "votes": null
    },
    {
      "id": "907596",
      "postDate": "06/30/2020 03:48:19",
      "content": "<p><a href=\"/zhenglv\">@zhenglv</a> zhengThank you for this tip. I was wondering how did you achieve that?</p>",
      "rawMarkdown": "zhenglv zhengThank you for this tip. I was wondering how did you achieve that?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 855493,
      "author_name": "zhenglv",
      "author_url": "",
      "post_date": "05/21/2020 01:43:21",
      "content": "<p>Solved by reducing batch size.</p>",
      "votes": null,
      "replies": [
        {
          "id": 905445,
          "author_name": "tonychenxyz",
          "author_url": "",
          "post_date": "06/28/2020 14:56:22",
          "content": "<p><a href=\"/zhenglv\">@zhenglv</a> What batch size did you set at the beginning, and what did you use to solve the problem? Thanks!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 905944,
      "author_name": "zhenglv",
      "author_url": "",
      "post_date": "06/29/2020 01:21:16",
      "content": "<p>64 * 8  failed , 32 * 8  succeeded</p>",
      "votes": null,
      "replies": [
        {
          "id": 905957,
          "author_name": "tonychenxyz",
          "author_url": "",
          "post_date": "06/29/2020 01:55:29",
          "content": "<p><a href=\"/zhenglv\">@zhenglv</a> That's weird. I even have the error when having 8 * 8 in colab. solved the error by removing .cache(), but training went very slow that way.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 907450,
      "author_name": "zhenglv",
      "author_url": "",
      "post_date": "06/30/2020 01:11:38",
      "content": "<p>Have you changed the pics into tfrecords ? After changing them into tfrecords , my note runs a epoch about 10 mins.</p>",
      "votes": null,
      "replies": [
        {
          "id": 907596,
          "author_name": "tonychenxyz",
          "author_url": "",
          "post_date": "06/30/2020 03:48:19",
          "content": "<p><a href=\"/zhenglv\">@zhenglv</a> zhengThank you for this tip. I was wondering how did you achieve that?</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "854694": "when using tpu training, at the end of the first epoch, met this error, any idea?\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F4044110%2F292c5af5778cbc3af8828972b49b25e1%2Ftpu_error.png?generation=1589963588974029&amp;alt=media)",
    "855493": "Solved by reducing batch size.",
    "905445": "zhenglv What batch size did you set at the beginning, and what did you use to solve the problem? Thanks!",
    "905944": "64 * 8  failed , 32 * 8  succeeded",
    "905957": "zhenglv That's weird. I even have the error when having 8 * 8 in colab. solved the error by removing .cache(), but training went very slow that way.",
    "907450": "Have you changed the pics into tfrecords ? After changing them into tfrecords , my note runs a epoch about 10 mins.",
    "907596": "zhenglv zhengThank you for this tip. I was wondering how did you achieve that?"
  },
  "source": "meta"
}