{
  "id": 125494,
  "title": "Notebook Threw Exception",
  "url": "/competitions/bengaliai-cv19/discussion/125494",
  "author_name": "",
  "post_date": "2020-01-11T05:54:23.866512700Z",
  "votes": 2,
  "comment_count": 5,
  "views": 0,
  "content": "<p>When I run my inference code in the notebook, it runs fine (even when I use train dataset so the size is similar to the actual submission run). But when i submit, it's always Notebook Threw Exception. Has anyone else run into this error? I've been trying to debug it for a week now. My inference code is <a href=\"https://www.kaggle.com/shujun717/01072019?scriptVersionId=26672365\">https://www.kaggle.com/shujun717/01072019?scriptVersionId=26672365</a></p>",
  "messages": [
    {
      "id": "715995",
      "postDate": "01/11/2020 05:54:23",
      "content": "<p>When I run my inference code in the notebook, it runs fine (even when I use train dataset so the size is similar to the actual submission run). But when i submit, it's always Notebook Threw Exception. Has anyone else run into this error? I've been trying to debug it for a week now. My inference code is <a href=\"https://www.kaggle.com/shujun717/01072019?scriptVersionId=26672365\">https://www.kaggle.com/shujun717/01072019?scriptVersionId=26672365</a></p>",
      "rawMarkdown": "When I run my inference code in the notebook, it runs fine (even when I use train dataset so the size is similar to the actual submission run). But when i submit, it's always Notebook Threw Exception. Has anyone else run into this error? I've been trying to debug it for a week now. My inference code is https://www.kaggle.com/shujun717/01072019?scriptVersionId=26672365",
      "votes": null
    },
    {
      "id": "716044",
      "postDate": "01/11/2020 07:15:06",
      "content": "<p>Hi shujun,</p>\n\n<p>Try removing box and crop_resize function as these function consumes lot of memory.</p>\n\n<p>Try resizing with just cv2.resize() these will help you removing unwanted code. </p>\n\n<p>I am sure you will get lead from this point.</p>",
      "rawMarkdown": "Hi shujun,\n\nTry removing box and crop_resize function as these function consumes lot of memory.\n\nTry resizing with just cv2.resize() these will help you removing unwanted code. \n\nI am sure you will get lead from this point.",
      "votes": null
    },
    {
      "id": "716045",
      "postDate": "01/11/2020 07:16:01",
      "content": "<p>Will do. Thanks for the advice!</p>",
      "rawMarkdown": "Will do. Thanks for the advice!",
      "votes": null
    },
    {
      "id": "717366",
      "postDate": "01/13/2020 03:58:56",
      "content": "<p>Hey do you know why those functions consume a lot of memory? I thought variables in functions get released after the function finishes?</p>",
      "rawMarkdown": "Hey do you know why those functions consume a lot of memory? I thought variables in functions get released after the function finishes?",
      "votes": null
    },
    {
      "id": "717375",
      "postDate": "01/13/2020 04:08:01",
      "content": "<p>Hey Shujun did this solved your problem?</p>\n\n<p>I realized that these functions are not releasing the memory even after computation is finished.  It keeps on piling up on every call leading to OOM error.\nAlso this memory is not recovered by GC.collect as it went out of scope.</p>",
      "rawMarkdown": "Hey Shujun did this solved your problem?\n\nI realized that these functions are not releasing the memory even after computation is finished.  It keeps on piling up on every call leading to OOM error.\nAlso this memory is not recovered by GC.collect as it went out of scope.",
      "votes": null
    },
    {
      "id": "717438",
      "postDate": "01/13/2020 05:29:36",
      "content": "<p>It did help solve my problem so much thanks! I was able to train new models with just resize and got a score of 0.9613. However, i feel like the crop_resize data preprocessing could help my accuracy so I want to find a way to fix the memory issue. Do you know if it can be fixed?</p>",
      "rawMarkdown": "It did help solve my problem so much thanks! I was able to train new models with just resize and got a score of 0.9613. However, i feel like the crop_resize data preprocessing could help my accuracy so I want to find a way to fix the memory issue. Do you know if it can be fixed?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 716044,
      "author_name": "amit9484",
      "author_url": "",
      "post_date": "01/11/2020 07:15:06",
      "content": "<p>Hi shujun,</p>\n\n<p>Try removing box and crop_resize function as these function consumes lot of memory.</p>\n\n<p>Try resizing with just cv2.resize() these will help you removing unwanted code. </p>\n\n<p>I am sure you will get lead from this point.</p>",
      "votes": null,
      "replies": [
        {
          "id": 716045,
          "author_name": "shujun717",
          "author_url": "",
          "post_date": "01/11/2020 07:16:01",
          "content": "<p>Will do. Thanks for the advice!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 717366,
          "author_name": "shujun717",
          "author_url": "",
          "post_date": "01/13/2020 03:58:56",
          "content": "<p>Hey do you know why those functions consume a lot of memory? I thought variables in functions get released after the function finishes?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 717375,
          "author_name": "amit9484",
          "author_url": "",
          "post_date": "01/13/2020 04:08:01",
          "content": "<p>Hey Shujun did this solved your problem?</p>\n\n<p>I realized that these functions are not releasing the memory even after computation is finished.  It keeps on piling up on every call leading to OOM error.\nAlso this memory is not recovered by GC.collect as it went out of scope.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 717438,
          "author_name": "shujun717",
          "author_url": "",
          "post_date": "01/13/2020 05:29:36",
          "content": "<p>It did help solve my problem so much thanks! I was able to train new models with just resize and got a score of 0.9613. However, i feel like the crop_resize data preprocessing could help my accuracy so I want to find a way to fix the memory issue. Do you know if it can be fixed?</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "715995": "When I run my inference code in the notebook, it runs fine (even when I use train dataset so the size is similar to the actual submission run). But when i submit, it's always Notebook Threw Exception. Has anyone else run into this error? I've been trying to debug it for a week now. My inference code is https://www.kaggle.com/shujun717/01072019?scriptVersionId=26672365",
    "716044": "Hi shujun,\n\nTry removing box and crop_resize function as these function consumes lot of memory.\n\nTry resizing with just cv2.resize() these will help you removing unwanted code. \n\nI am sure you will get lead from this point.",
    "716045": "Will do. Thanks for the advice!",
    "717366": "Hey do you know why those functions consume a lot of memory? I thought variables in functions get released after the function finishes?",
    "717375": "Hey Shujun did this solved your problem?\n\nI realized that these functions are not releasing the memory even after computation is finished.  It keeps on piling up on every call leading to OOM error.\nAlso this memory is not recovered by GC.collect as it went out of scope.",
    "717438": "It did help solve my problem so much thanks! I was able to train new models with just resize and got a score of 0.9613. However, i feel like the crop_resize data preprocessing could help my accuracy so I want to find a way to fix the memory issue. Do you know if it can be fixed?"
  },
  "source": "meta"
}