{
  "id": 583375,
  "title": "Is the training code based on Egor Trushin's float16 UNet notebook?",
  "url": "/competitions/waveform-inversion/discussion/583375",
  "author_name": "CodeHacker",
  "post_date": "2025-06-06T11:47:30.613000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi everyone,<br>\nI'm exploring the training pipelines used in this competition and came across a previous UNet-based approach using float16.</p>\n<p>I'd like to know whether the current best-performing training code (e.g., shared in top discussions or kernels) is a modified version of that baseline, or if it has evolved into something completely different.</p>\n<p>Thanks in advance for any insights!</p>",
  "messages": [
    {
      "id": 3218583,
      "postDate": "2025-06-06T11:47:30.613Z",
      "content": "<p>Hi everyone,<br>\nI'm exploring the training pipelines used in this competition and came across a previous UNet-based approach using float16.</p>\n<p>I'd like to know whether the current best-performing training code (e.g., shared in top discussions or kernels) is a modified version of that baseline, or if it has evolved into something completely different.</p>\n<p>Thanks in advance for any insights!</p>",
      "rawMarkdown": "Hi everyone,\nI'm exploring the training pipelines used in this competition and came across a previous UNet-based approach using float16.\n\nI'd like to know whether the current best-performing training code (e.g., shared in top discussions or kernels) is a modified version of that baseline, or if it has evolved into something completely different.\n\nThanks in advance for any insights!",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3218583": "Hi everyone,\nI'm exploring the training pipelines used in this competition and came across a previous UNet-based approach using float16.\n\nI'd like to know whether the current best-performing training code (e.g., shared in top discussions or kernels) is a modified version of that baseline, or if it has evolved into something completely different.\n\nThanks in advance for any insights!"
  }
}