{
  "id": 104099,
  "title": "Mean gets affected while converting scalar_coupling_constant to float16",
  "url": "/competitions/champs-scalar-coupling/discussion/104099",
  "author_name": "",
  "post_date": "2019-08-14T08:58:27.129699400Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi I am using the function that reduces the memory usage mentioned in one of the kernels but while doing EDA, I found out that mean printed was NaN. I searched and found out that float16 is not supported by pandas.</p>\n\n<p>See here:\n<a href=\"https://github.com/pandas-dev/pandas/issues/20642\">https://github.com/pandas-dev/pandas/issues/20642</a></p>",
  "messages": [
    {
      "id": "598940",
      "postDate": "08/14/2019 08:58:27",
      "content": "<p>Hi I am using the function that reduces the memory usage mentioned in one of the kernels but while doing EDA, I found out that mean printed was NaN. I searched and found out that float16 is not supported by pandas.</p>\n\n<p>See here:\n<a href=\"https://github.com/pandas-dev/pandas/issues/20642\">https://github.com/pandas-dev/pandas/issues/20642</a></p>",
      "rawMarkdown": "Hi I am using the function that reduces the memory usage mentioned in one of the kernels but while doing EDA, I found out that mean printed was NaN. I searched and found out that float16 is not supported by pandas.\n\nSee here:\nhttps://github.com/pandas-dev/pandas/issues/20642",
      "votes": null
    },
    {
      "id": "600790",
      "postDate": "08/16/2019 15:01:34",
      "content": "<p>Hi, I only have 12 giga byte of ram, but I don't calculate features with float 16. You can have side effects.</p>\n\n<p>You can calculate the features with larger floats and then load them as float 16 to train the model. :-)</p>",
      "rawMarkdown": "Hi, I only have 12 giga byte of ram, but I don't calculate features with float 16. You can have side effects.\n\nYou can calculate the features with larger floats and then load them as float 16 to train the model. :-)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 600790,
      "author_name": "raphaelsourty",
      "author_url": "",
      "post_date": "08/16/2019 15:01:34",
      "content": "<p>Hi, I only have 12 giga byte of ram, but I don't calculate features with float 16. You can have side effects.</p>\n\n<p>You can calculate the features with larger floats and then load them as float 16 to train the model. :-)</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "598940": "Hi I am using the function that reduces the memory usage mentioned in one of the kernels but while doing EDA, I found out that mean printed was NaN. I searched and found out that float16 is not supported by pandas.\n\nSee here:\nhttps://github.com/pandas-dev/pandas/issues/20642",
    "600790": "Hi, I only have 12 giga byte of ram, but I don't calculate features with float 16. You can have side effects.\n\nYou can calculate the features with larger floats and then load them as float 16 to train the model. :-)"
  },
  "source": "meta"
}