{
  "id": 109681,
  "title": "Submission taking 20+ minutes to score",
  "url": "/competitions/open-images-2019-instance-segmentation/discussion/109681",
  "author_name": "",
  "post_date": "2019-09-21T08:46:25.638590900Z",
  "votes": 1,
  "comment_count": 7,
  "views": 0,
  "content": "<p>Submitted a 77Mb file for scoring and it's been over 20 mins and I haven't gotten a score. \nDue to this I'm not able to submit any other predictions as well :/</p>\n\n<p>In the past when I submitted files with larger sizes (500Mb+) it used to take only 2-3 mins to score. </p>\n\n<p>Kaggle Team, can you please look into this? </p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1168459%2F82d1e5cc7d099b30de67a8ef98e87114%2Fkaggle_submission_error.png?generation=1569055446025256&amp;alt=media\" alt=\"\"></p>\n\n<p>Edit : <a href=\"/juliaelliott\">@juliaelliott</a> <a href=\"/philculliton\">@philculliton</a> </p>",
  "messages": [
    {
      "id": "631054",
      "postDate": "09/21/2019 08:46:25",
      "content": "<p>Submitted a 77Mb file for scoring and it's been over 20 mins and I haven't gotten a score. \nDue to this I'm not able to submit any other predictions as well :/</p>\n\n<p>In the past when I submitted files with larger sizes (500Mb+) it used to take only 2-3 mins to score. </p>\n\n<p>Kaggle Team, can you please look into this? </p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1168459%2F82d1e5cc7d099b30de67a8ef98e87114%2Fkaggle_submission_error.png?generation=1569055446025256&amp;alt=media\" alt=\"\"></p>\n\n<p>Edit : <a href=\"/juliaelliott\">@juliaelliott</a> <a href=\"/philculliton\">@philculliton</a> </p>",
      "rawMarkdown": "Submitted a 77Mb file for scoring and it's been over 20 mins and I haven't gotten a score. \nDue to this I'm not able to submit any other predictions as well :/\n\nIn the past when I submitted files with larger sizes (500Mb+) it used to take only 2-3 mins to score. \n\nKaggle Team, can you please look into this? \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1168459%2F82d1e5cc7d099b30de67a8ef98e87114%2Fkaggle_submission_error.png?generation=1569055446025256&amp;alt=media)\n\nEdit : @juliaelliott @philculliton",
      "votes": null
    },
    {
      "id": "634692",
      "postDate": "09/26/2019 16:14:02",
      "content": "<p>Hi <a href=\"/rohitmidha23\">@rohitmidha23</a> - sorry I missed this!  Thanks for the report.  We're taking a look.</p>",
      "rawMarkdown": "Hi @rohitmidha23 - sorry I missed this!  Thanks for the report.  We're taking a look.",
      "votes": null
    },
    {
      "id": "634702",
      "postDate": "09/26/2019 16:31:51",
      "content": "<p>Checked this out - looks like it might have been an issue with server load at the time you were submitting, as I haven't been able to reproduce it. The submissions completes in a few seconds when I run it. Have you had continuing issues with this?</p>",
      "rawMarkdown": "Checked this out - looks like it might have been an issue with server load at the time you were submitting, as I haven't been able to reproduce it. The submissions completes in a few seconds when I run it. Have you had continuing issues with this?",
      "votes": null
    },
    {
      "id": "634739",
      "postDate": "09/26/2019 17:25:50",
      "content": "<p><a href=\"/philculliton\">@philculliton</a> thanks for replying. Since then I've been able to make submissions, yes. No issues right now :smile:</p>",
      "rawMarkdown": "philculliton thanks for replying. Since then I've been able to make submissions, yes. No issues right now :smile:",
      "votes": null
    },
    {
      "id": "634936",
      "postDate": "09/27/2019 01:07:43",
      "content": "<p><a href=\"/philculliton\">@philculliton</a> Hi, whether it is possible to tell me how the kaggle evaluates the submission quickly. The script on the github implemented by tensorflow seems slowly. Thanks.</p>",
      "rawMarkdown": "philculliton Hi, whether it is possible to tell me how the kaggle evaluates the submission quickly. The script on the github implemented by tensorflow seems slowly. Thanks.",
      "votes": null
    },
    {
      "id": "634938",
      "postDate": "09/27/2019 01:11:14",
      "content": "<p><a href=\"/rohitmidha23\">@rohitmidha23</a> - that's great to hear!  :-)  Let me know if any more issues come up.</p>",
      "rawMarkdown": "rohitmidha23 - that's great to hear!  :-)  Let me know if any more issues come up.",
      "votes": null
    },
    {
      "id": "634940",
      "postDate": "09/27/2019 01:12:53",
      "content": "<p>Hi <a href=\"/wangchuanyuan\">@wangchuanyuan</a> - the Kaggle back end uses heavily optimized C# code based on the Python code.  It is specialized for the competition and runs on many cores. We do a lot of work on the back end to make sure that code runs as quickly as possible!</p>",
      "rawMarkdown": "Hi @wangchuanyuan - the Kaggle back end uses heavily optimized C# code based on the Python code.  It is specialized for the competition and runs on many cores. We do a lot of work on the back end to make sure that code runs as quickly as possible!",
      "votes": null
    },
    {
      "id": "634946",
      "postDate": "09/27/2019 01:26:52",
      "content": "<p>Thanks for your excellent work 👍. It seems the np.nonzero operation to transform the mask into bounding box costs a lot. Could you please give me some suggestions? Or I should remove it and use the predicted bounding box directly.</p>",
      "rawMarkdown": "Thanks for your excellent work 👍. It seems the np.nonzero operation to transform the mask into bounding box costs a lot. Could you please give me some suggestions? Or I should remove it and use the predicted bounding box directly.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 634692,
      "author_name": "philculliton",
      "author_url": "",
      "post_date": "09/26/2019 16:14:02",
      "content": "<p>Hi <a href=\"/rohitmidha23\">@rohitmidha23</a> - sorry I missed this!  Thanks for the report.  We're taking a look.</p>",
      "votes": null,
      "replies": [
        {
          "id": 634702,
          "author_name": "philculliton",
          "author_url": "",
          "post_date": "09/26/2019 16:31:51",
          "content": "<p>Checked this out - looks like it might have been an issue with server load at the time you were submitting, as I haven't been able to reproduce it. The submissions completes in a few seconds when I run it. Have you had continuing issues with this?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 634739,
          "author_name": "rohitmidha23",
          "author_url": "",
          "post_date": "09/26/2019 17:25:50",
          "content": "<p><a href=\"/philculliton\">@philculliton</a> thanks for replying. Since then I've been able to make submissions, yes. No issues right now :smile:</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 634938,
          "author_name": "philculliton",
          "author_url": "",
          "post_date": "09/27/2019 01:11:14",
          "content": "<p><a href=\"/rohitmidha23\">@rohitmidha23</a> - that's great to hear!  :-)  Let me know if any more issues come up.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 634936,
      "author_name": "wangchuanyuan",
      "author_url": "",
      "post_date": "09/27/2019 01:07:43",
      "content": "<p><a href=\"/philculliton\">@philculliton</a> Hi, whether it is possible to tell me how the kaggle evaluates the submission quickly. The script on the github implemented by tensorflow seems slowly. Thanks.</p>",
      "votes": null,
      "replies": [
        {
          "id": 634940,
          "author_name": "philculliton",
          "author_url": "",
          "post_date": "09/27/2019 01:12:53",
          "content": "<p>Hi <a href=\"/wangchuanyuan\">@wangchuanyuan</a> - the Kaggle back end uses heavily optimized C# code based on the Python code.  It is specialized for the competition and runs on many cores. We do a lot of work on the back end to make sure that code runs as quickly as possible!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 634946,
          "author_name": "wangchuanyuan",
          "author_url": "",
          "post_date": "09/27/2019 01:26:52",
          "content": "<p>Thanks for your excellent work 👍. It seems the np.nonzero operation to transform the mask into bounding box costs a lot. Could you please give me some suggestions? Or I should remove it and use the predicted bounding box directly.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "631054": "Submitted a 77Mb file for scoring and it's been over 20 mins and I haven't gotten a score. \nDue to this I'm not able to submit any other predictions as well :/\n\nIn the past when I submitted files with larger sizes (500Mb+) it used to take only 2-3 mins to score. \n\nKaggle Team, can you please look into this? \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1168459%2F82d1e5cc7d099b30de67a8ef98e87114%2Fkaggle_submission_error.png?generation=1569055446025256&amp;alt=media)\n\nEdit : @juliaelliott @philculliton",
    "634692": "Hi @rohitmidha23 - sorry I missed this!  Thanks for the report.  We're taking a look.",
    "634702": "Checked this out - looks like it might have been an issue with server load at the time you were submitting, as I haven't been able to reproduce it. The submissions completes in a few seconds when I run it. Have you had continuing issues with this?",
    "634739": "philculliton thanks for replying. Since then I've been able to make submissions, yes. No issues right now :smile:",
    "634936": "philculliton Hi, whether it is possible to tell me how the kaggle evaluates the submission quickly. The script on the github implemented by tensorflow seems slowly. Thanks.",
    "634938": "rohitmidha23 - that's great to hear!  :-)  Let me know if any more issues come up.",
    "634940": "Hi @wangchuanyuan - the Kaggle back end uses heavily optimized C# code based on the Python code.  It is specialized for the competition and runs on many cores. We do a lot of work on the back end to make sure that code runs as quickly as possible!",
    "634946": "Thanks for your excellent work 👍. It seems the np.nonzero operation to transform the mask into bounding box costs a lot. Could you please give me some suggestions? Or I should remove it and use the predicted bounding box directly."
  },
  "source": "meta"
}