{
  "id": 131428,
  "title": "process of submission",
  "url": "/competitions/bengaliai-cv19/discussion/131428",
  "author_name": "",
  "post_date": "2020-02-19T20:37:48.679847100Z",
  "votes": null,
  "comment_count": 7,
  "views": 0,
  "content": "<p>The process of submission on Bengali.Ai take a lot of time . I have more than 4 hours waiting the result of my submission, but the form submission always show that the submission is ongoing. Please give help to fix this issue.</p>",
  "messages": [
    {
      "id": "750924",
      "postDate": "02/19/2020 20:37:48",
      "content": "<p>The process of submission on Bengali.Ai take a lot of time . I have more than 4 hours waiting the result of my submission, but the form submission always show that the submission is ongoing. Please give help to fix this issue.</p>",
      "rawMarkdown": "The process of submission on Bengali.Ai take a lot of time . I have more than 4 hours waiting the result of my submission, but the form submission always show that the submission is ongoing. Please give help to fix this issue.",
      "votes": null
    },
    {
      "id": "751052",
      "postDate": "02/20/2020 00:47:00",
      "content": "<p>Your submitted code must predict labels for 200,000 test images! This takes time. You can check how long this takes by predicting labels on all the train images while you sit and watch. That is how long your submission will take. If that is over 2 hours, you must speed it up.</p>",
      "rawMarkdown": "Your submitted code must predict labels for 200,000 test images! This takes time. You can check how long this takes by predicting labels on all the train images while you sit and watch. That is how long your submission will take. If that is over 2 hours, you must speed it up.",
      "votes": null
    },
    {
      "id": "751053",
      "postDate": "02/20/2020 00:47:19",
      "content": "<p>What is your batch size? it depends on model performance. My previous model was taking about 45mins with 256 batch size but the current model takes 17mins with 128 batch size.</p>",
      "rawMarkdown": "What is your batch size? it depends on model performance. My previous model was taking about 45mins with 256 batch size but the current model takes 17mins with 128 batch size.",
      "votes": null
    },
    {
      "id": "751139",
      "postDate": "02/20/2020 02:46:46",
      "content": "<p>Hey Chris, how do you know there are 200,000 test images? Thank you.</p>",
      "rawMarkdown": "Hey Chris, how do you know there are 200,000 test images? Thank you.",
      "votes": null
    },
    {
      "id": "751359",
      "postDate": "02/20/2020 06:46:29",
      "content": "<p>Thank you Chris for your reply. My submission take more then 4 hours than it give an error message of timeout . Besides i wonder that how did you find that there are 200.000 test images. Could you allow us grasp this. Thank you .</p>",
      "rawMarkdown": "Thank you Chris for your reply. My submission take more then 4 hours than it give an error message of timeout . Besides i wonder that how did you find that there are 200.000 test images. Could you allow us grasp this. Thank you .",
      "votes": null
    },
    {
      "id": "751363",
      "postDate": "02/20/2020 06:51:32",
      "content": "<p>Hi Corey and Taher. The competition host said (in a reply <a href=\"https://www.kaggle.com/c/bengaliai-cv19/discussion/123135#704272\">here</a>) that the number of train and test are similar. Furthermore, the time it takes for my models to infer all the training is the same time it takes my submission to return an LB score.</p>",
      "rawMarkdown": "Hi Corey and Taher. The competition host said (in a reply [here][1]) that the number of train and test are similar. Furthermore, the time it takes for my models to infer all the training is the same time it takes my submission to return an LB score.\n\n[1]: https://www.kaggle.com/c/bengaliai-cv19/discussion/123135#704272",
      "votes": null
    },
    {
      "id": "751371",
      "postDate": "02/20/2020 06:54:03",
      "content": "<p><a href=\"/taherhaggui\">@taherhaggui</a> This is what I suggest you do. Open an interactive notebook. Then run your inference code on the 200,000 train images. Watch your notebook and make sure it predicts all the train images without having any memory issues and finishes under 2 hours. If both are true then it will submit fine (when you change the word \"train\" that references the parquets to \"test\").</p>",
      "rawMarkdown": "taherhaggui This is what I suggest you do. Open an interactive notebook. Then run your inference code on the 200,000 train images. Watch your notebook and make sure it predicts all the train images without having any memory issues and finishes under 2 hours. If both are true then it will submit fine (when you change the word \"train\" that references the parquets to \"test\").",
      "votes": null
    },
    {
      "id": "751441",
      "postDate": "02/20/2020 07:42:17",
      "content": "<p><a href=\"/cdeotte\">@cdeotte</a>  Thank you Chris again for the accurate reply . It is strange for me that my submision take more than 4 hours to be finalised then it either give an error message for the timeout or it give a score very strange. For my case it give me a score less than 1% when i have a score in my validation data aound 96% . We should notice that time of my model training is about 2 hours , I finished the my GPU hours  and my submission file encomapasses 36 rows correspond to 12 images test . Please give me an explanation , i'm lost.</p>",
      "rawMarkdown": "cdeotte  Thank you Chris again for the accurate reply . It is strange for me that my submision take more than 4 hours to be finalised then it either give an error message for the timeout or it give a score very strange. For my case it give me a score less than 1% when i have a score in my validation data aound 96% . We should notice that time of my model training is about 2 hours , I finished the my GPU hours  and my submission file encomapasses 36 rows correspond to 12 images test . Please give me an explanation , i'm lost.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 751052,
      "author_name": "cdeotte",
      "author_url": "",
      "post_date": "02/20/2020 00:47:00",
      "content": "<p>Your submitted code must predict labels for 200,000 test images! This takes time. You can check how long this takes by predicting labels on all the train images while you sit and watch. That is how long your submission will take. If that is over 2 hours, you must speed it up.</p>",
      "votes": null,
      "replies": [
        {
          "id": 751139,
          "author_name": "returnofsputnik",
          "author_url": "",
          "post_date": "02/20/2020 02:46:46",
          "content": "<p>Hey Chris, how do you know there are 200,000 test images? Thank you.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 751359,
          "author_name": "taherhaggui",
          "author_url": "",
          "post_date": "02/20/2020 06:46:29",
          "content": "<p>Thank you Chris for your reply. My submission take more then 4 hours than it give an error message of timeout . Besides i wonder that how did you find that there are 200.000 test images. Could you allow us grasp this. Thank you .</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 751363,
          "author_name": "cdeotte",
          "author_url": "",
          "post_date": "02/20/2020 06:51:32",
          "content": "<p>Hi Corey and Taher. The competition host said (in a reply <a href=\"https://www.kaggle.com/c/bengaliai-cv19/discussion/123135#704272\">here</a>) that the number of train and test are similar. Furthermore, the time it takes for my models to infer all the training is the same time it takes my submission to return an LB score.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 751371,
          "author_name": "cdeotte",
          "author_url": "",
          "post_date": "02/20/2020 06:54:03",
          "content": "<p><a href=\"/taherhaggui\">@taherhaggui</a> This is what I suggest you do. Open an interactive notebook. Then run your inference code on the 200,000 train images. Watch your notebook and make sure it predicts all the train images without having any memory issues and finishes under 2 hours. If both are true then it will submit fine (when you change the word \"train\" that references the parquets to \"test\").</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 751441,
          "author_name": "taherhaggui",
          "author_url": "",
          "post_date": "02/20/2020 07:42:17",
          "content": "<p><a href=\"/cdeotte\">@cdeotte</a>  Thank you Chris again for the accurate reply . It is strange for me that my submision take more than 4 hours to be finalised then it either give an error message for the timeout or it give a score very strange. For my case it give me a score less than 1% when i have a score in my validation data aound 96% . We should notice that time of my model training is about 2 hours , I finished the my GPU hours  and my submission file encomapasses 36 rows correspond to 12 images test . Please give me an explanation , i'm lost.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 751053,
      "author_name": "shamrat3628",
      "author_url": "",
      "post_date": "02/20/2020 00:47:19",
      "content": "<p>What is your batch size? it depends on model performance. My previous model was taking about 45mins with 256 batch size but the current model takes 17mins with 128 batch size.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "750924": "The process of submission on Bengali.Ai take a lot of time . I have more than 4 hours waiting the result of my submission, but the form submission always show that the submission is ongoing. Please give help to fix this issue.",
    "751052": "Your submitted code must predict labels for 200,000 test images! This takes time. You can check how long this takes by predicting labels on all the train images while you sit and watch. That is how long your submission will take. If that is over 2 hours, you must speed it up.",
    "751053": "What is your batch size? it depends on model performance. My previous model was taking about 45mins with 256 batch size but the current model takes 17mins with 128 batch size.",
    "751139": "Hey Chris, how do you know there are 200,000 test images? Thank you.",
    "751359": "Thank you Chris for your reply. My submission take more then 4 hours than it give an error message of timeout . Besides i wonder that how did you find that there are 200.000 test images. Could you allow us grasp this. Thank you .",
    "751363": "Hi Corey and Taher. The competition host said (in a reply [here][1]) that the number of train and test are similar. Furthermore, the time it takes for my models to infer all the training is the same time it takes my submission to return an LB score.\n\n[1]: https://www.kaggle.com/c/bengaliai-cv19/discussion/123135#704272",
    "751371": "taherhaggui This is what I suggest you do. Open an interactive notebook. Then run your inference code on the 200,000 train images. Watch your notebook and make sure it predicts all the train images without having any memory issues and finishes under 2 hours. If both are true then it will submit fine (when you change the word \"train\" that references the parquets to \"test\").",
    "751441": "cdeotte  Thank you Chris again for the accurate reply . It is strange for me that my submision take more than 4 hours to be finalised then it either give an error message for the timeout or it give a score very strange. For my case it give me a score less than 1% when i have a score in my validation data aound 96% . We should notice that time of my model training is about 2 hours , I finished the my GPU hours  and my submission file encomapasses 36 rows correspond to 12 images test . Please give me an explanation , i'm lost."
  },
  "source": "meta"
}