{
  "id": 140131,
  "title": "Combining both datasets on TPU",
  "url": "/competitions/jigsaw-multilingual-toxic-comment-classification/discussion/140131",
  "author_name": "DimitreOliveira",
  "post_date": "2020-03-31T14:11:34.100000",
  "votes": 3,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi everyone, I was running on some issues while trying to combine both datasets and run on TPU, some issues were discussed on <a href=\"https://www.kaggle.com/c/jigsaw-multilingual-toxic-comment-classification/discussion/140007\">this thread</a>, but the solution as pointed by  @qinhui1999 , is that to run data on TPU the labels needs to be on the same type, so basically to combine both datasets and run on TPU you must cast both labels to <code>int</code> or <code>float</code>.</p>",
  "messages": [
    {
      "id": 792726,
      "postDate": "2020-03-31T14:11:34.100Z",
      "content": "<p>Hi everyone, I was running on some issues while trying to combine both datasets and run on TPU, some issues were discussed on <a href=\"https://www.kaggle.com/c/jigsaw-multilingual-toxic-comment-classification/discussion/140007\">this thread</a>, but the solution as pointed by  @qinhui1999 , is that to run data on TPU the labels needs to be on the same type, so basically to combine both datasets and run on TPU you must cast both labels to <code>int</code> or <code>float</code>.</p>",
      "rawMarkdown": "Hi everyone, I was running on some issues while trying to combine both datasets and run on TPU, some issues were discussed on [this thread](https://www.kaggle.com/c/jigsaw-multilingual-toxic-comment-classification/discussion/140007), but the solution as pointed by  @qinhui1999 , is that to run data on TPU the labels needs to be on the same type, so basically to combine both datasets and run on TPU you must cast both labels to `int` or `float`.",
      "votes": 3
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "792726": "Hi everyone, I was running on some issues while trying to combine both datasets and run on TPU, some issues were discussed on [this thread](https://www.kaggle.com/c/jigsaw-multilingual-toxic-comment-classification/discussion/140007), but the solution as pointed by  @qinhui1999 , is that to run data on TPU the labels needs to be on the same type, so basically to combine both datasets and run on TPU you must cast both labels to `int` or `float`."
  }
}