{
  "id": 205626,
  "title": "How to Implement SAKT with Tensorflow?",
  "url": "/competitions/riiid-test-answer-prediction/discussion/205626",
  "author_name": "",
  "post_date": "2020-12-21T02:31:09.799429400Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi Kagglers,</p>\n<p>I am new to this competition,<br>\nI've read lots of public notebooks mostly people using LGB and SAKT models.<br>\nSAKT model structure looks like 3 embedding layers 2 with one data and 1 with other data. and passing 2 embedding outputs to the attention model. and then drop out and 2FC layers.</p>\n<p>if I want to build the same structure with TensorFlow how can I build it?<br>\nhow can I build 2 Embedding layers and feed them to the attention layer with TensorFlow?</p>\n<p>Thanks,</p>\n<p>Vishal</p>",
  "messages": [
    {
      "id": "1120646",
      "postDate": "12/21/2020 02:31:09",
      "content": "<p>Hi Kagglers,</p>\n<p>I am new to this competition,<br>\nI've read lots of public notebooks mostly people using LGB and SAKT models.<br>\nSAKT model structure looks like 3 embedding layers 2 with one data and 1 with other data. and passing 2 embedding outputs to the attention model. and then drop out and 2FC layers.</p>\n<p>if I want to build the same structure with TensorFlow how can I build it?<br>\nhow can I build 2 Embedding layers and feed them to the attention layer with TensorFlow?</p>\n<p>Thanks,</p>\n<p>Vishal</p>",
      "rawMarkdown": "Hi Kagglers,\n\nI am new to this competition,\nI've read lots of public notebooks mostly people using LGB and SAKT models.\nSAKT model structure looks like 3 embedding layers 2 with one data and 1 with other data. and passing 2 embedding outputs to the attention model. and then drop out and 2FC layers.\n\nif I want to build the same structure with TensorFlow how can I build it?\nhow can I build 2 Embedding layers and feed them to the attention layer with TensorFlow?\n\nThanks,\n\nVishal",
      "votes": null
    },
    {
      "id": "1120728",
      "postDate": "12/21/2020 04:45:39",
      "content": "<p><a href=\"https://github.com/shalini1194/SAKT\" target=\"_blank\">https://github.com/shalini1194/SAKT</a></p>",
      "rawMarkdown": "https://github.com/shalini1194/SAKT",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1120728,
      "author_name": "projdev",
      "author_url": "",
      "post_date": "12/21/2020 04:45:39",
      "content": "<p><a href=\"https://github.com/shalini1194/SAKT\" target=\"_blank\">https://github.com/shalini1194/SAKT</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1120646": "Hi Kagglers,\n\nI am new to this competition,\nI've read lots of public notebooks mostly people using LGB and SAKT models.\nSAKT model structure looks like 3 embedding layers 2 with one data and 1 with other data. and passing 2 embedding outputs to the attention model. and then drop out and 2FC layers.\n\nif I want to build the same structure with TensorFlow how can I build it?\nhow can I build 2 Embedding layers and feed them to the attention layer with TensorFlow?\n\nThanks,\n\nVishal",
    "1120728": "https://github.com/shalini1194/SAKT"
  },
  "source": "meta"
}