{
  "id": 108300,
  "title": "how to step-by-step train",
  "url": "/competitions/open-images-2019-object-detection/discussion/108300",
  "author_name": "",
  "post_date": "2019-09-10T15:23:14.654652600Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Currently I use mobilenet in tf hub to train this model, but I want to use faster r-cnn to train my model. However, this is beyond the kaggle's GPU usage time, so I would like to ask how to carry out step-by-step training so that I can complete this game?\n😁</p>",
  "messages": [
    {
      "id": "623185",
      "postDate": "09/10/2019 15:23:14",
      "content": "<p>Currently I use mobilenet in tf hub to train this model, but I want to use faster r-cnn to train my model. However, this is beyond the kaggle's GPU usage time, so I would like to ask how to carry out step-by-step training so that I can complete this game?\n😁</p>",
      "rawMarkdown": "Currently I use mobilenet in tf hub to train this model, but I want to use faster r-cnn to train my model. However, this is beyond the kaggle's GPU usage time, so I would like to ask how to carry out step-by-step training so that I can complete this game?\n😁",
      "votes": null
    },
    {
      "id": "623366",
      "postDate": "09/10/2019 20:09:36",
      "content": "<p><a href=\"https://www.kaggle.com/akashdeepjassal/tf-hub-inception-resnet\">It is done here.</a> That said, you need another model with comparable performance, and that is where things get tricky, because it takes days and days to train something comparable even with a formidable GPU. They were giving away GCP credits for this competition for reason. Frankly, I haven't started serious work on this competition until I was notified that I got my GCP credit. Otherwise, it would take a month to do anything worth doing with my GTX 1070. </p>",
      "rawMarkdown": "[It is done here.](https://www.kaggle.com/akashdeepjassal/tf-hub-inception-resnet) That said, you need another model with comparable performance, and that is where things get tricky, because it takes days and days to train something comparable even with a formidable GPU. They were giving away GCP credits for this competition for reason. Frankly, I haven't started serious work on this competition until I was notified that I got my GCP credit. Otherwise, it would take a month to do anything worth doing with my GTX 1070.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 623366,
      "author_name": "dzspin",
      "author_url": "",
      "post_date": "09/10/2019 20:09:36",
      "content": "<p><a href=\"https://www.kaggle.com/akashdeepjassal/tf-hub-inception-resnet\">It is done here.</a> That said, you need another model with comparable performance, and that is where things get tricky, because it takes days and days to train something comparable even with a formidable GPU. They were giving away GCP credits for this competition for reason. Frankly, I haven't started serious work on this competition until I was notified that I got my GCP credit. Otherwise, it would take a month to do anything worth doing with my GTX 1070. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "623185": "Currently I use mobilenet in tf hub to train this model, but I want to use faster r-cnn to train my model. However, this is beyond the kaggle's GPU usage time, so I would like to ask how to carry out step-by-step training so that I can complete this game?\n😁",
    "623366": "[It is done here.](https://www.kaggle.com/akashdeepjassal/tf-hub-inception-resnet) That said, you need another model with comparable performance, and that is where things get tricky, because it takes days and days to train something comparable even with a formidable GPU. They were giving away GCP credits for this competition for reason. Frankly, I haven't started serious work on this competition until I was notified that I got my GCP credit. Otherwise, it would take a month to do anything worth doing with my GTX 1070."
  },
  "source": "meta"
}