{
  "id": 161674,
  "title": "Confusion about the feature formats of tfrecord files.",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/161674",
  "author_name": "",
  "post_date": "2020-06-25T16:57:14.606462Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi Kaggle,\nI am a beginner in Kaggle. How do you know about the feature formats of the tfrecord files? I could not get it from the Data section of the competition.</p>",
  "messages": [
    {
      "id": "901743",
      "postDate": "06/25/2020 16:57:14",
      "content": "<p>Hi Kaggle,\nI am a beginner in Kaggle. How do you know about the feature formats of the tfrecord files? I could not get it from the Data section of the competition.</p>",
      "rawMarkdown": "Hi Kaggle,\nI am a beginner in Kaggle. How do you know about the feature formats of the tfrecord files? I could not get it from the Data section of the competition.",
      "votes": null
    },
    {
      "id": "901761",
      "postDate": "06/25/2020 17:16:27",
      "content": "<p>hey <a href=\"/umongsain\">@umongsain</a>  , \nyou can view the data section there are tfrecords available to be easily used for faster training on either GPU or TPU which ever you prefer, but TPU is really very fast. \nand for feature formats ,if you are not able to understand it properly , then you  create your own tfrecords for this purpose. For better understanding on this you can view this discussion  <a href=\"https://www.kaggle.com/c/siim-isic-melanoma-classification/discussion/155579\">https://www.kaggle.com/c/siim-isic-melanoma-classification/discussion/155579</a> , a great resource from <a href=\"/cdeotte\">@cdeotte</a> . Just have a look once , it surely helpful to you.</p>",
      "rawMarkdown": "hey @umongsain  , \nyou can view the data section there are tfrecords available to be easily used for faster training on either GPU or TPU which ever you prefer, but TPU is really very fast. \nand for feature formats ,if you are not able to understand it properly , then you  create your own tfrecords for this purpose. For better understanding on this you can view this discussion  https://www.kaggle.com/c/siim-isic-melanoma-classification/discussion/155579 , a great resource from @cdeotte . Just have a look once , it surely helpful to you.",
      "votes": null
    },
    {
      "id": "902425",
      "postDate": "06/26/2020 06:22:10",
      "content": "<p>Thanks a lot, <a href=\"/prashantarora\">@prashantarora</a> 😀.  But again, why didn't the competition organizers say anything clearly about the fields of the tfrecord files?</p>",
      "rawMarkdown": "Thanks a lot, @prashantarora 😀.  But again, why didn't the competition organizers say anything clearly about the fields of the tfrecord files?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 901761,
      "author_name": "prashantarorat",
      "author_url": "",
      "post_date": "06/25/2020 17:16:27",
      "content": "<p>hey <a href=\"/umongsain\">@umongsain</a>  , \nyou can view the data section there are tfrecords available to be easily used for faster training on either GPU or TPU which ever you prefer, but TPU is really very fast. \nand for feature formats ,if you are not able to understand it properly , then you  create your own tfrecords for this purpose. For better understanding on this you can view this discussion  <a href=\"https://www.kaggle.com/c/siim-isic-melanoma-classification/discussion/155579\">https://www.kaggle.com/c/siim-isic-melanoma-classification/discussion/155579</a> , a great resource from <a href=\"/cdeotte\">@cdeotte</a> . Just have a look once , it surely helpful to you.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 902425,
      "author_name": "umongsain",
      "author_url": "",
      "post_date": "06/26/2020 06:22:10",
      "content": "<p>Thanks a lot, <a href=\"/prashantarora\">@prashantarora</a> 😀.  But again, why didn't the competition organizers say anything clearly about the fields of the tfrecord files?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "901743": "Hi Kaggle,\nI am a beginner in Kaggle. How do you know about the feature formats of the tfrecord files? I could not get it from the Data section of the competition.",
    "901761": "hey @umongsain  , \nyou can view the data section there are tfrecords available to be easily used for faster training on either GPU or TPU which ever you prefer, but TPU is really very fast. \nand for feature formats ,if you are not able to understand it properly , then you  create your own tfrecords for this purpose. For better understanding on this you can view this discussion  https://www.kaggle.com/c/siim-isic-melanoma-classification/discussion/155579 , a great resource from @cdeotte . Just have a look once , it surely helpful to you.",
    "902425": "Thanks a lot, @prashantarora 😀.  But again, why didn't the competition organizers say anything clearly about the fields of the tfrecord files?"
  },
  "source": "meta"
}