{
  "id": 409384,
  "title": "baseline comment(for new kaggler)",
  "url": "/competitions/vesuvius-challenge-ink-detection/discussion/409384",
  "author_name": "",
  "post_date": "2023-05-10T18:18:35.870674100Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I added a more detailed comment to the baseline, I hope you point out if there is a mistake in understanding<br>\n<a href=\"https://www.kaggle.com/code/allenwpr/baseline-comment-for-new-kaggler\" target=\"_blank\">https://www.kaggle.com/code/allenwpr/baseline-comment-for-new-kaggler</a></p>",
  "messages": [
    {
      "id": "2254216",
      "postDate": "05/10/2023 18:18:35",
      "content": "<p>I added a more detailed comment to the baseline, I hope you point out if there is a mistake in understanding<br>\n<a href=\"https://www.kaggle.com/code/allenwpr/baseline-comment-for-new-kaggler\" target=\"_blank\">https://www.kaggle.com/code/allenwpr/baseline-comment-for-new-kaggler</a></p>",
      "rawMarkdown": "I added a more detailed comment to the baseline, I hope you point out if there is a mistake in understanding\nhttps://www.kaggle.com/code/allenwpr/baseline-comment-for-new-kaggler",
      "votes": null
    },
    {
      "id": "2254839",
      "postDate": "05/11/2023 09:24:15",
      "content": "<p>I think that your comment on the use of GradScaler in the context of autocast is largely correct. Still, I would emphasize that the reason for using these tools together is to address the numerical instability issues caused by the limited range of representable values of float16.</p>",
      "rawMarkdown": "I think that your comment on the use of GradScaler in the context of autocast is largely correct. Still, I would emphasize that the reason for using these tools together is to address the numerical instability issues caused by the limited range of representable values of float16.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2254839,
      "author_name": "quattrinifabio",
      "author_url": "",
      "post_date": "05/11/2023 09:24:15",
      "content": "<p>I think that your comment on the use of GradScaler in the context of autocast is largely correct. Still, I would emphasize that the reason for using these tools together is to address the numerical instability issues caused by the limited range of representable values of float16.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2254216": "I added a more detailed comment to the baseline, I hope you point out if there is a mistake in understanding\nhttps://www.kaggle.com/code/allenwpr/baseline-comment-for-new-kaggler",
    "2254839": "I think that your comment on the use of GradScaler in the context of autocast is largely correct. Still, I would emphasize that the reason for using these tools together is to address the numerical instability issues caused by the limited range of representable values of float16."
  },
  "source": "meta"
}