{
  "id": 125606,
  "title": "Validation Evaluation During Training",
  "url": "/competitions/pku-autonomous-driving/discussion/125606",
  "author_name": "",
  "post_date": "2020-01-12T07:30:03.798164Z",
  "votes": 1,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi!</p>\n\n<p>Do you use validation loss or validation mAP in training process to determine model saving or stopping training?</p>",
  "messages": [
    {
      "id": "716742",
      "postDate": "01/12/2020 07:30:03",
      "content": "<p>Hi!</p>\n\n<p>Do you use validation loss or validation mAP in training process to determine model saving or stopping training?</p>",
      "rawMarkdown": "Hi!\n\nDo you use validation loss or validation mAP in training process to determine model saving or stopping training?",
      "votes": null
    },
    {
      "id": "716743",
      "postDate": "01/12/2020 07:30:37",
      "content": "<p>I wonder if using mAP would have different results from val loss.</p>",
      "rawMarkdown": "I wonder if using mAP would have different results from val loss.",
      "votes": null
    },
    {
      "id": "716748",
      "postDate": "01/12/2020 07:42:47",
      "content": "<p>I don't know what deep-learning framework you use, but if you are using tensorflow/keras, it is pretty easy to do that by customizing your own callback function.</p>",
      "rawMarkdown": "I don't know what deep-learning framework you use, but if you are using tensorflow/keras, it is pretty easy to do that by customizing your own callback function.",
      "votes": null
    },
    {
      "id": "717082",
      "postDate": "01/12/2020 17:44:07",
      "content": "<p><a href=\"/xiejialun\">@xiejialun</a>  I know how to implement callback. My question was which one did you use to determine model checkpoints. </p>",
      "rawMarkdown": "xiejialun  I know how to implement callback. My question was which one did you use to determine model checkpoints.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 716743,
      "author_name": "tonychenxyz",
      "author_url": "",
      "post_date": "01/12/2020 07:30:37",
      "content": "<p>I wonder if using mAP would have different results from val loss.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 716748,
      "author_name": "xiejialun",
      "author_url": "",
      "post_date": "01/12/2020 07:42:47",
      "content": "<p>I don't know what deep-learning framework you use, but if you are using tensorflow/keras, it is pretty easy to do that by customizing your own callback function.</p>",
      "votes": null,
      "replies": [
        {
          "id": 717082,
          "author_name": "tonychenxyz",
          "author_url": "",
          "post_date": "01/12/2020 17:44:07",
          "content": "<p><a href=\"/xiejialun\">@xiejialun</a>  I know how to implement callback. My question was which one did you use to determine model checkpoints. </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "716742": "Hi!\n\nDo you use validation loss or validation mAP in training process to determine model saving or stopping training?",
    "716743": "I wonder if using mAP would have different results from val loss.",
    "716748": "I don't know what deep-learning framework you use, but if you are using tensorflow/keras, it is pretty easy to do that by customizing your own callback function.",
    "717082": "xiejialun  I know how to implement callback. My question was which one did you use to determine model checkpoints."
  },
  "source": "meta"
}