{
  "id": 578149,
  "title": "Anyway to use flash-attention 2 in kaggle notebooks?",
  "url": "/competitions/stanford-rna-3d-folding/discussion/578149",
  "author_name": "Simon Beck",
  "post_date": "2025-05-09T04:34:05.617000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I'm currently trying to use flash-attention 2.3.2 in order to use RNA language model with <strong>GPU P100</strong><br>\nbut get <br>\nRuntimeError: FlashAttention only supports Ampere GPUs or newer.</p>\n<p>Is there any way to use it on kaggle notebooks?</p>\n<p>if not, is it possible for competition hosts to allow submitting inference result too?<br>\nwould really love to test out specific model im using</p>",
  "messages": [
    {
      "id": 3198108,
      "postDate": "2025-05-09T04:34:05.617Z",
      "content": "<p>I'm currently trying to use flash-attention 2.3.2 in order to use RNA language model with <strong>GPU P100</strong><br>\nbut get <br>\nRuntimeError: FlashAttention only supports Ampere GPUs or newer.</p>\n<p>Is there any way to use it on kaggle notebooks?</p>\n<p>if not, is it possible for competition hosts to allow submitting inference result too?<br>\nwould really love to test out specific model im using</p>",
      "rawMarkdown": "I'm currently trying to use flash-attention 2.3.2 in order to use RNA language model with **GPU P100**\nbut get \nRuntimeError: FlashAttention only supports Ampere GPUs or newer.\n\nIs there any way to use it on kaggle notebooks?\n\nif not, is it possible for competition hosts to allow submitting inference result too?\nwould really love to test out specific model im using",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3198108": "I'm currently trying to use flash-attention 2.3.2 in order to use RNA language model with **GPU P100**\nbut get \nRuntimeError: FlashAttention only supports Ampere GPUs or newer.\n\nIs there any way to use it on kaggle notebooks?\n\nif not, is it possible for competition hosts to allow submitting inference result too?\nwould really love to test out specific model im using"
  }
}