{
  "id": 70819,
  "title": "Can we have apex added to the gpu docker image?",
  "url": "/competitions/quora-insincere-questions-classification/discussion/70819",
  "author_name": "bilal2vec",
  "post_date": "2018-11-07T14:44:49.513000",
  "votes": 2,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi,</p>\n\n<p>Apex, (<a href=\"https://github.com/NVIDIA/apex\">https://github.com/NVIDIA/apex</a>) is a library by nvidia which makes training in mixed precision in Pytorch easy. Since this competition has a 2 hour limit on gpu kernels, mixed precision training will allow larger models and batch sizes to be used. </p>\n\n<p>Is it possible to have it added to the default gpu docker image since this competition doesn't allow submitting from custom docker images?</p>",
  "messages": [
    {
      "id": 416969,
      "postDate": "2018-11-07T14:44:49.513Z",
      "content": "<p>Hi,</p>\n\n<p>Apex, (<a href=\"https://github.com/NVIDIA/apex\">https://github.com/NVIDIA/apex</a>) is a library by nvidia which makes training in mixed precision in Pytorch easy. Since this competition has a 2 hour limit on gpu kernels, mixed precision training will allow larger models and batch sizes to be used. </p>\n\n<p>Is it possible to have it added to the default gpu docker image since this competition doesn't allow submitting from custom docker images?</p>",
      "rawMarkdown": "Hi,\n\nApex, ([https://github.com/NVIDIA/apex][1]) is a library by nvidia which makes training in mixed precision in Pytorch easy. Since this competition has a 2 hour limit on gpu kernels, mixed precision training will allow larger models and batch sizes to be used. \n\nIs it possible to have it added to the default gpu docker image since this competition doesn't allow submitting from custom docker images?\n\n  [1]: https://github.com/NVIDIA/apex",
      "votes": 2
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "416969": "Hi,\n\nApex, ([https://github.com/NVIDIA/apex][1]) is a library by nvidia which makes training in mixed precision in Pytorch easy. Since this competition has a 2 hour limit on gpu kernels, mixed precision training will allow larger models and batch sizes to be used. \n\nIs it possible to have it added to the default gpu docker image since this competition doesn't allow submitting from custom docker images?\n\n  [1]: https://github.com/NVIDIA/apex"
  }
}