{
  "id": 371401,
  "title": "Having trouble in the inference part",
  "url": "/competitions/rsna-breast-cancer-detection/discussion/371401",
  "author_name": "Myo Min Htet(wnp)🇲🇲",
  "post_date": "2022-12-09T17:40:20.422000",
  "votes": 0,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I got the outputs from the model and how do I make these outputs and make submission. According to the test set, I got four prediction tensor outputs and then  put into a list using cpu().detach().numpy(). We make a threshold and compare to all items in that list. And finally make a submission csv. Is that right? I don't know how to get threshold. My score is 0 even though roc_auc and f1 score is good in training part.</p>",
  "messages": [
    {
      "id": 2061151,
      "postDate": "2022-12-10T18:24:19.170Z",
      "content": "<p>Check out some public notebooks under <code>Code</code> in which contributors also show how they are doing inference e.g.: <a href=\"https://www.kaggle.com/code/theoviel/rsna-breast-baseline-inference\" target=\"_blank\">https://www.kaggle.com/code/theoviel/rsna-breast-baseline-inference</a></p>",
      "rawMarkdown": "Check out some public notebooks under `Code` in which contributors also show how they are doing inference e.g.: https://www.kaggle.com/code/theoviel/rsna-breast-baseline-inference"
    },
    {
      "id": 2060293,
      "postDate": "2022-12-09T17:40:20.423Z",
      "content": "<p>I got the outputs from the model and how do I make these outputs and make submission. According to the test set, I got four prediction tensor outputs and then  put into a list using cpu().detach().numpy(). We make a threshold and compare to all items in that list. And finally make a submission csv. Is that right? I don't know how to get threshold. My score is 0 even though roc_auc and f1 score is good in training part.</p>",
      "rawMarkdown": "I got the outputs from the model and how do I make these outputs and make submission. According to the test set, I got four prediction tensor outputs and then  put into a list using cpu().detach().numpy(). We make a threshold and compare to all items in that list. And finally make a submission csv. Is that right? I don't know how to get threshold. My score is 0 even though roc_auc and f1 score is good in training part."
    }
  ],
  "comments": [
    {
      "id": 2061151,
      "author_name": "Ali Abdin",
      "author_url": "",
      "post_date": "2022-12-10T18:24:19.170000",
      "content": "<p>Check out some public notebooks under <code>Code</code> in which contributors also show how they are doing inference e.g.: <a href=\"https://www.kaggle.com/code/theoviel/rsna-breast-baseline-inference\" target=\"_blank\">https://www.kaggle.com/code/theoviel/rsna-breast-baseline-inference</a></p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2061151": "Check out some public notebooks under `Code` in which contributors also show how they are doing inference e.g.: https://www.kaggle.com/code/theoviel/rsna-breast-baseline-inference",
    "2060293": "I got the outputs from the model and how do I make these outputs and make submission. According to the test set, I got four prediction tensor outputs and then  put into a list using cpu().detach().numpy(). We make a threshold and compare to all items in that list. And finally make a submission csv. Is that right? I don't know how to get threshold. My score is 0 even though roc_auc and f1 score is good in training part."
  }
}