{
  "id": 283854,
  "title": " I Implemented Nfnets-keras from scratch and got silver on Shopee - Try them out!",
  "url": "/competitions/wikipedia-image-caption/discussion/283854",
  "author_name": "",
  "post_date": "2021-10-28T07:13:40.195000",
  "votes": 4,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hey guys,</p>\n<p>A couple of months ago I implemented nfnets for Keras from scratch for the Shopee competition [there wasn't any at the time.. 😅].</p>\n<p>Although similar competitions from the past, shoppe is probably the closest one we got when searching for refrences. </p>\n<p>Anyway, my team's solution was based on the nfnets and ended with <strong>silver</strong>.<br>\n[There was also a vision transformer for keras I also implemented from scratch]</p>\n<p><strong>If you are interested, check them out here:</strong><br>\n<strong><a href=\"https://github.com/ypeleg/nfnets-keras\" target=\"_blank\">https://github.com/ypeleg/nfnets-keras</a></strong></p>\n<p>It has been a while since I played with them so there might be some compatibility issues but overall they should be working seamlessly.</p>\n<p>I might release the vision transformer however, there are many current keras implementations for vision transformers so there is no real point besides getting feedback for my own code..</p>\n<p>See you in a month,<br>\nHappy Kaggling!</p>\n<p>Cheers^^</p>",
  "messages": [
    {
      "id": 1563297,
      "postDate": "2021-10-28T07:13:40.197Z",
      "content": "<p>Hey guys,</p>\n<p>A couple of months ago I implemented nfnets for Keras from scratch for the Shopee competition [there wasn't any at the time.. 😅].</p>\n<p>Although similar competitions from the past, shoppe is probably the closest one we got when searching for refrences. </p>\n<p>Anyway, my team's solution was based on the nfnets and ended with <strong>silver</strong>.<br>\n[There was also a vision transformer for keras I also implemented from scratch]</p>\n<p><strong>If you are interested, check them out here:</strong><br>\n<strong><a href=\"https://github.com/ypeleg/nfnets-keras\" target=\"_blank\">https://github.com/ypeleg/nfnets-keras</a></strong></p>\n<p>It has been a while since I played with them so there might be some compatibility issues but overall they should be working seamlessly.</p>\n<p>I might release the vision transformer however, there are many current keras implementations for vision transformers so there is no real point besides getting feedback for my own code..</p>\n<p>See you in a month,<br>\nHappy Kaggling!</p>\n<p>Cheers^^</p>",
      "rawMarkdown": "Hey guys,\n\nA couple of months ago I implemented nfnets for Keras from scratch for the Shopee competition [there wasn't any at the time.. 😅].\n\nAlthough similar competitions from the past, shoppe is probably the closest one we got when searching for refrences. \n\nAnyway, my team's solution was based on the nfnets and ended with **silver**.\n[There was also a vision transformer for keras I also implemented from scratch]\n\n**If you are interested, check them out here:**\n**https://github.com/ypeleg/nfnets-keras**\n\nIt has been a while since I played with them so there might be some compatibility issues but overall they should be working seamlessly.\n\nI might release the vision transformer however, there are many current keras implementations for vision transformers so there is no real point besides getting feedback for my own code..\n\nSee you in a month,\nHappy Kaggling!\n\nCheers^^",
      "votes": 4
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1563297": "Hey guys,\n\nA couple of months ago I implemented nfnets for Keras from scratch for the Shopee competition [there wasn't any at the time.. 😅].\n\nAlthough similar competitions from the past, shoppe is probably the closest one we got when searching for refrences. \n\nAnyway, my team's solution was based on the nfnets and ended with **silver**.\n[There was also a vision transformer for keras I also implemented from scratch]\n\n**If you are interested, check them out here:**\n**https://github.com/ypeleg/nfnets-keras**\n\nIt has been a while since I played with them so there might be some compatibility issues but overall they should be working seamlessly.\n\nI might release the vision transformer however, there are many current keras implementations for vision transformers so there is no real point besides getting feedback for my own code..\n\nSee you in a month,\nHappy Kaggling!\n\nCheers^^"
  }
}