{
  "id": 157648,
  "title": "kernel died after 3 folds - [TPU]",
  "url": "/competitions/jigsaw-multilingual-toxic-comment-classification/discussion/157648",
  "author_name": "",
  "post_date": "2020-06-11T14:05:08.619275500Z",
  "votes": 2,
  "comment_count": 4,
  "views": 0,
  "content": "<p>I am trying to build 5 fold model using XLMRoberta(Large), But for some reason my kernel is getting killed after 3 Folds(continuously) without any resource Execution warning/error. \nfacing this issue even after using <code>tf.tpu.experimental.initialize_tpu_system(tpu)</code> before starting each fold.\nAny suggestions?</p>",
  "messages": [
    {
      "id": "881966",
      "postDate": "06/11/2020 14:05:08",
      "content": "<p>I am trying to build 5 fold model using XLMRoberta(Large), But for some reason my kernel is getting killed after 3 Folds(continuously) without any resource Execution warning/error. \nfacing this issue even after using <code>tf.tpu.experimental.initialize_tpu_system(tpu)</code> before starting each fold.\nAny suggestions?</p>",
      "rawMarkdown": "I am trying to build 5 fold model using XLMRoberta(Large), But for some reason my kernel is getting killed after 3 Folds(continuously) without any resource Execution warning/error. \nfacing this issue even after using ` tf.tpu.experimental.initialize_tpu_system(tpu)` before starting each fold.\nAny suggestions?",
      "votes": null
    },
    {
      "id": "888131",
      "postDate": "06/16/2020 06:41:25",
      "content": "<p>reduce the train_batch_size and valid_batch_size and try.</p>",
      "rawMarkdown": "reduce the train_batch_size and valid_batch_size and try.",
      "votes": null
    },
    {
      "id": "888320",
      "postDate": "06/16/2020 09:16:21",
      "content": "<p>Probably you are hitting 3 hours limit.</p>",
      "rawMarkdown": "Probably you are hitting 3 hours limit.",
      "votes": null
    },
    {
      "id": "888873",
      "postDate": "06/16/2020 16:02:32",
      "content": "<p>You are right. my kernel reached 3 hours limit</p>",
      "rawMarkdown": "You are right. my kernel reached 3 hours limit",
      "votes": null
    },
    {
      "id": "890244",
      "postDate": "06/17/2020 12:04:43",
      "content": "<p>good to know you know what was the reason. You can use colab if you want as an alternative. </p>",
      "rawMarkdown": "good to know you know what was the reason. You can use colab if you want as an alternative.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 888131,
      "author_name": "shivyshiv",
      "author_url": "",
      "post_date": "06/16/2020 06:41:25",
      "content": "<p>reduce the train_batch_size and valid_batch_size and try.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 888320,
      "author_name": "urvishp80",
      "author_url": "",
      "post_date": "06/16/2020 09:16:21",
      "content": "<p>Probably you are hitting 3 hours limit.</p>",
      "votes": null,
      "replies": [
        {
          "id": 888873,
          "author_name": "ashoksrinivas",
          "author_url": "",
          "post_date": "06/16/2020 16:02:32",
          "content": "<p>You are right. my kernel reached 3 hours limit</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 890244,
          "author_name": "urvishp80",
          "author_url": "",
          "post_date": "06/17/2020 12:04:43",
          "content": "<p>good to know you know what was the reason. You can use colab if you want as an alternative. </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "881966": "I am trying to build 5 fold model using XLMRoberta(Large), But for some reason my kernel is getting killed after 3 Folds(continuously) without any resource Execution warning/error. \nfacing this issue even after using ` tf.tpu.experimental.initialize_tpu_system(tpu)` before starting each fold.\nAny suggestions?",
    "888131": "reduce the train_batch_size and valid_batch_size and try.",
    "888320": "Probably you are hitting 3 hours limit.",
    "888873": "You are right. my kernel reached 3 hours limit",
    "890244": "good to know you know what was the reason. You can use colab if you want as an alternative."
  },
  "source": "meta"
}