{
  "id": 232754,
  "title": "Could you share something about your experience with ViT or DeiT?",
  "url": "/competitions/ranzcr-clip-catheter-line-classification/discussion/232754",
  "author_name": "Jerry_Jiang",
  "post_date": "2021-04-15T08:29:54.446000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I am trying to using the vit and DeiT for this task, but i got a poor valid CV score. I used the end-to-end training model instead of the pre training model. The image size is set as 448, and batch size is set as 32. Unfortunately, the Roc is slowwing down during the training.  I really have no idea about that. I am first time to use transformer for CV task. Could you share something about your experience with ViT or DeiT? Thanks a lot!🙏🙏</p>",
  "messages": [
    {
      "id": 1274364,
      "postDate": "2021-04-15T08:29:54.447Z",
      "content": "<p>I am trying to using the vit and DeiT for this task, but i got a poor valid CV score. I used the end-to-end training model instead of the pre training model. The image size is set as 448, and batch size is set as 32. Unfortunately, the Roc is slowwing down during the training.  I really have no idea about that. I am first time to use transformer for CV task. Could you share something about your experience with ViT or DeiT? Thanks a lot!🙏🙏</p>",
      "rawMarkdown": "I am trying to using the vit and DeiT for this task, but i got a poor valid CV score. I used the end-to-end training model instead of the pre training model. The image size is set as 448, and batch size is set as 32. Unfortunately, the Roc is slowwing down during the training.  I really have no idea about that. I am first time to use transformer for CV task. Could you share something about your experience with ViT or DeiT? Thanks a lot!🙏🙏",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1274364": "I am trying to using the vit and DeiT for this task, but i got a poor valid CV score. I used the end-to-end training model instead of the pre training model. The image size is set as 448, and batch size is set as 32. Unfortunately, the Roc is slowwing down during the training.  I really have no idea about that. I am first time to use transformer for CV task. Could you share something about your experience with ViT or DeiT? Thanks a lot!🙏🙏"
  }
}