{
  "id": 97620,
  "title": "The kernel appears to have died. It will restart automatically",
  "url": "/competitions/aptos2019-blindness-detection/discussion/97620",
  "author_name": "",
  "post_date": "2019-06-28T03:30:44.165783100Z",
  "votes": 7,
  "comment_count": 9,
  "views": 0,
  "content": "<p>God... really.... you are killing me here.......3 times already\nI am either kind of dumb... or this kernel thing is still unstable.</p>",
  "messages": [
    {
      "id": "563225",
      "postDate": "06/28/2019 03:30:44",
      "content": "<p>God... really.... you are killing me here.......3 times already\nI am either kind of dumb... or this kernel thing is still unstable.</p>",
      "rawMarkdown": "God... really.... you are killing me here.......3 times already\nI am either kind of dumb... or this kernel thing is still unstable.",
      "votes": null
    },
    {
      "id": "563955",
      "postDate": "06/28/2019 20:16:39",
      "content": "<p>Same here !</p>",
      "rawMarkdown": "Same here !",
      "votes": null
    },
    {
      "id": "563978",
      "postDate": "06/28/2019 20:57:24",
      "content": "<p>I'm guessing too little RAM. Kaggle should have more RAM in the KO competitions. More often than not kernels die from running out of memory (OOM). <code>del</code> and <code>gc.collect()</code> are my best friends in these cases.</p>",
      "rawMarkdown": "I'm guessing too little RAM. Kaggle should have more RAM in the KO competitions. More often than not kernels die from running out of memory (OOM). `del` and `gc.collect()` are my best friends in these cases.",
      "votes": null
    },
    {
      "id": "564993",
      "postDate": "06/30/2019 09:37:39",
      "content": "<p>Thanks for sharing </p>",
      "rawMarkdown": "Thanks for sharing",
      "votes": null
    },
    {
      "id": "565058",
      "postDate": "06/30/2019 11:24:46",
      "content": "<p>As others said, it can be because of insufficient RAM.  There is no warning nor error message in that case.</p>",
      "rawMarkdown": "As others said, it can be because of insufficient RAM.  There is no warning nor error message in that case.",
      "votes": null
    },
    {
      "id": "565553",
      "postDate": "07/01/2019 05:26:52",
      "content": "<p>I had this on <code>Predicting Molecular Properties</code> while using XGBoost GPU. On CPU it worked though but was so long !\nNot sure it was memory in that case.</p>",
      "rawMarkdown": "I had this on `Predicting Molecular Properties` while using XGBoost GPU. On CPU it worked though but was so long !\nNot sure it was memory in that case.",
      "votes": null
    },
    {
      "id": "565935",
      "postDate": "07/01/2019 15:23:40",
      "content": "<p>Totally unstable, browser to browser, time to time, GPU to CPU, Edit or creation,  interruption is killing me , l built lots of models locally, but rarely can build a new kernel or edit previous one.  </p>",
      "rawMarkdown": "Totally unstable, browser to browser, time to time, GPU to CPU, Edit or creation,  interruption is killing me , l built lots of models locally, but rarely can build a new kernel or edit previous one.",
      "votes": null
    },
    {
      "id": "566822",
      "postDate": "07/02/2019 16:25:59",
      "content": "<p>GPUs are bit unstable in Kaggle. Used to reconnect much frequently nowadays. If you have RAM limitations then you can try switching to CPU and also delete the variables that are not needed. </p>",
      "rawMarkdown": "GPUs are bit unstable in Kaggle. Used to reconnect much frequently nowadays. If you have RAM limitations then you can try switching to CPU and also delete the variables that are not needed.",
      "votes": null
    },
    {
      "id": "569671",
      "postDate": "07/07/2019 06:01:55",
      "content": "<p>I keep facing the same problem again and again. Unable to try out different methods because the kernel dies off most of the time after reading in the data.</p>",
      "rawMarkdown": "I keep facing the same problem again and again. Unable to try out different methods because the kernel dies off most of the time after reading in the data.",
      "votes": null
    },
    {
      "id": "587996",
      "postDate": "07/30/2019 02:28:26",
      "content": "<p>I have the same issue and I tried turning off the GPU, but it still doesn't work the kernel keeps dying</p>",
      "rawMarkdown": "I have the same issue and I tried turning off the GPU, but it still doesn't work the kernel keeps dying",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 563955,
      "author_name": "arashnic",
      "author_url": "",
      "post_date": "06/28/2019 20:16:39",
      "content": "<p>Same here !</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 563978,
      "author_name": "teeyee314",
      "author_url": "",
      "post_date": "06/28/2019 20:57:24",
      "content": "<p>I'm guessing too little RAM. Kaggle should have more RAM in the KO competitions. More often than not kernels die from running out of memory (OOM). <code>del</code> and <code>gc.collect()</code> are my best friends in these cases.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 564993,
      "author_name": "madhur321",
      "author_url": "",
      "post_date": "06/30/2019 09:37:39",
      "content": "<p>Thanks for sharing </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 565058,
      "author_name": "cpmpml",
      "author_url": "",
      "post_date": "06/30/2019 11:24:46",
      "content": "<p>As others said, it can be because of insufficient RAM.  There is no warning nor error message in that case.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 565553,
      "author_name": "ogrellier",
      "author_url": "",
      "post_date": "07/01/2019 05:26:52",
      "content": "<p>I had this on <code>Predicting Molecular Properties</code> while using XGBoost GPU. On CPU it worked though but was so long !\nNot sure it was memory in that case.</p>",
      "votes": null,
      "replies": [
        {
          "id": 565935,
          "author_name": "arashnic",
          "author_url": "",
          "post_date": "07/01/2019 15:23:40",
          "content": "<p>Totally unstable, browser to browser, time to time, GPU to CPU, Edit or creation,  interruption is killing me , l built lots of models locally, but rarely can build a new kernel or edit previous one.  </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 587996,
          "author_name": "luizfelipebjcosta",
          "author_url": "",
          "post_date": "07/30/2019 02:28:26",
          "content": "<p>I have the same issue and I tried turning off the GPU, but it still doesn't work the kernel keeps dying</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 566822,
      "author_name": "venkateshprabhug",
      "author_url": "",
      "post_date": "07/02/2019 16:25:59",
      "content": "<p>GPUs are bit unstable in Kaggle. Used to reconnect much frequently nowadays. If you have RAM limitations then you can try switching to CPU and also delete the variables that are not needed. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 569671,
      "author_name": "rhtsingh",
      "author_url": "",
      "post_date": "07/07/2019 06:01:55",
      "content": "<p>I keep facing the same problem again and again. Unable to try out different methods because the kernel dies off most of the time after reading in the data.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "563225": "God... really.... you are killing me here.......3 times already\nI am either kind of dumb... or this kernel thing is still unstable.",
    "563955": "Same here !",
    "563978": "I'm guessing too little RAM. Kaggle should have more RAM in the KO competitions. More often than not kernels die from running out of memory (OOM). `del` and `gc.collect()` are my best friends in these cases.",
    "564993": "Thanks for sharing",
    "565058": "As others said, it can be because of insufficient RAM.  There is no warning nor error message in that case.",
    "565553": "I had this on `Predicting Molecular Properties` while using XGBoost GPU. On CPU it worked though but was so long !\nNot sure it was memory in that case.",
    "565935": "Totally unstable, browser to browser, time to time, GPU to CPU, Edit or creation,  interruption is killing me , l built lots of models locally, but rarely can build a new kernel or edit previous one.",
    "566822": "GPUs are bit unstable in Kaggle. Used to reconnect much frequently nowadays. If you have RAM limitations then you can try switching to CPU and also delete the variables that are not needed.",
    "569671": "I keep facing the same problem again and again. Unable to try out different methods because the kernel dies off most of the time after reading in the data.",
    "587996": "I have the same issue and I tried turning off the GPU, but it still doesn't work the kernel keeps dying"
  },
  "source": "meta"
}