{
  "id": 327138,
  "title": "Parquet Format Dataset for Low Memory Use ",
  "url": "/competitions/amex-default-prediction/discussion/327138",
  "author_name": "Sanskar Hasija",
  "post_date": "2022-05-25T19:33:30.988000",
  "votes": 168,
  "comment_count": 70,
  "views": 0,
  "content": "<p>Hello, I have created Parquet format files for train_data.csv and test_data.csv files. The size of the CSV files decreased as follows:<br>\nFull Dataset: <b>50.31 GB</b> ➡️ <b>10.41 GB</b><br>\ntrain_data : <b>16.39 GB</b> ➡️ <b>3.35 GB</b><br>\ntest_data: <b>33.82 GB</b> ➡️ <b>6.96 GB</b></p>\n<p>Link to Dataset - <a href=\"http://www.kaggle.com/odins0n/amex-parquet\" target=\"_blank\">www.kaggle.com/odins0n/amex-parquet</a></p>\n<p>The columns with <b>float64</b> dtype were also converted to <b>float32</b> dtype as well.  Parquet files are much faster to read in for datasets of this size</p>\n<p>Example Notebook for loading Parquet files - <a href=\"http://www.kaggle.com/code/odins0n/load-parquet-files-with-low-memory/\" target=\"_blank\">www.kaggle.com/code/odins0n/load-parquet-files-with-low-memory/</a><br>\nPlease let me know if you have any feedback 😀</p>",
  "messages": [
    {
      "id": 1801471,
      "postDate": "2022-05-25T19:33:30.987Z",
      "content": "<p>Hello, I have created Parquet format files for train_data.csv and test_data.csv files. The size of the CSV files decreased as follows:<br>\nFull Dataset: <b>50.31 GB</b> ➡️ <b>10.41 GB</b><br>\ntrain_data : <b>16.39 GB</b> ➡️ <b>3.35 GB</b><br>\ntest_data: <b>33.82 GB</b> ➡️ <b>6.96 GB</b></p>\n<p>Link to Dataset - <a href=\"http://www.kaggle.com/odins0n/amex-parquet\" target=\"_blank\">www.kaggle.com/odins0n/amex-parquet</a></p>\n<p>The columns with <b>float64</b> dtype were also converted to <b>float32</b> dtype as well.  Parquet files are much faster to read in for datasets of this size</p>\n<p>Example Notebook for loading Parquet files - <a href=\"http://www.kaggle.com/code/odins0n/load-parquet-files-with-low-memory/\" target=\"_blank\">www.kaggle.com/code/odins0n/load-parquet-files-with-low-memory/</a><br>\nPlease let me know if you have any feedback 😀</p>",
      "rawMarkdown": "Hello, I have created Parquet format files for train_data.csv and test_data.csv files. The size of the CSV files decreased as follows:\nFull Dataset: <b>50.31 GB</b> ➡️ <b>10.41 GB</b>\ntrain_data : <b>16.39 GB</b> ➡️ <b>3.35 GB</b>\ntest_data: <b>33.82 GB</b> ➡️ <b>6.96 GB</b>\n\n\nLink to Dataset - www.kaggle.com/odins0n/amex-parquet\n\nThe columns with <b>float64</b> dtype were also converted to <b>float32</b> dtype as well.  Parquet files are much faster to read in for datasets of this size\n\nExample Notebook for loading Parquet files - www.kaggle.com/code/odins0n/load-parquet-files-with-low-memory/\nPlease let me know if you have any feedback 😀\n",
      "votes": 164
    },
    {
      "id": 1843525,
      "postDate": "2022-07-05T01:45:47.093Z",
      "content": "<p><img src=\"https://i.ibb.co/3hHnF5f/Untitled.png\" alt=\"\"></p>",
      "rawMarkdown": "![](https://i.ibb.co/3hHnF5f/Untitled.png)",
      "votes": 4
    },
    {
      "id": 1897923,
      "postDate": "2022-08-14T06:43:33.310Z",
      "content": "<p>that's very useful!</p>",
      "rawMarkdown": "that's very useful!",
      "votes": 1
    },
    {
      "id": 1833179,
      "postDate": "2022-06-25T17:48:06.860Z",
      "content": "<p>Thanks for sharing <a href=\"https://www.kaggle.com/odins0n\" target=\"_blank\">@odins0n</a>. Could you please tell how it was converted to parquet file?</p>",
      "rawMarkdown": "Thanks for sharing @odins0n. Could you please tell how it was converted to parquet file?",
      "votes": 1
    },
    {
      "id": 1807181,
      "postDate": "2022-05-31T19:46:25.737Z",
      "content": "<p>Thank you!! It would be very helpful!</p>",
      "rawMarkdown": "Thank you!! It would be very helpful!",
      "votes": 1,
      "replies": [
        {
          "id": 1809677,
          "postDate": "2022-06-03T00:08:56.447Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/michaelpoluektov\" target=\"_blank\">@michaelpoluektov</a> </p>",
          "rawMarkdown": "Welcome @michaelpoluektov "
        }
      ]
    },
    {
      "id": 1806783,
      "postDate": "2022-05-31T13:11:35.480Z",
      "content": "<p>Thanks for sharing! I noticed that <a href=\"https://www.kaggle.com/ruchi798\" target=\"_blank\">@ruchi798</a> has suggested using feather files with float16 <a href=\"https://www.kaggle.com/competitions/amex-default-prediction/discussion/327143\" target=\"_blank\">here</a>. I was wondering, since the values have been normalised and random noise has supposedly been added, do you think there are any advantages of using float32 instead of float16?</p>",
      "rawMarkdown": "Thanks for sharing! I noticed that @ruchi798 has suggested using feather files with float16 [here](https://www.kaggle.com/competitions/amex-default-prediction/discussion/327143). I was wondering, since the values have been normalised and random noise has supposedly been added, do you think there are any advantages of using float32 instead of float16?",
      "votes": 1,
      "replies": [
        {
          "id": 1809681,
          "postDate": "2022-06-03T00:11:29.777Z",
          "content": "<p>It will depend from feature to feature but in general yes. Float32 will perform better than float16 since more and more information will be lost when we compress it from the original float64 values. </p>",
          "rawMarkdown": "It will depend from feature to feature but in general yes. Float32 will perform better than float16 since more and more information will be lost when we compress it from the original float64 values. "
        }
      ]
    },
    {
      "id": 1806446,
      "postDate": "2022-05-31T07:13:22.830Z",
      "content": "<p>Thanks for sharing it</p>",
      "rawMarkdown": "Thanks for sharing it",
      "votes": 1,
      "replies": [
        {
          "id": 1809676,
          "postDate": "2022-06-03T00:08:43.110Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/imvision12\" target=\"_blank\">@imvision12</a> </p>",
          "rawMarkdown": "Welcome @imvision12 "
        }
      ]
    },
    {
      "id": 1806288,
      "postDate": "2022-05-31T02:38:12.457Z",
      "content": "<p>Thank you for sharing</p>",
      "rawMarkdown": "Thank you for sharing",
      "votes": 1,
      "replies": [
        {
          "id": 1806290,
          "postDate": "2022-05-31T02:42:46.403Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/pedrohsouzax\" target=\"_blank\">@pedrohsouzax</a> </p>",
          "rawMarkdown": "Welcome @pedrohsouzax "
        }
      ]
    },
    {
      "id": 1806258,
      "postDate": "2022-05-31T01:18:30.747Z",
      "content": "<p>Thanks for sharing brother</p>",
      "rawMarkdown": "Thanks for sharing brother",
      "votes": 1,
      "replies": [
        {
          "id": 1806259,
          "postDate": "2022-05-31T01:22:44.500Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/mungeunjo\" target=\"_blank\">@mungeunjo</a> </p>",
          "rawMarkdown": "Welcome @mungeunjo "
        }
      ]
    },
    {
      "id": 1806120,
      "postDate": "2022-05-30T19:57:11.817Z",
      "content": "<p>Thanks for this.</p>",
      "rawMarkdown": "Thanks for this.",
      "votes": 1,
      "replies": [
        {
          "id": 1806260,
          "postDate": "2022-05-31T01:22:51.340Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/abdurrakibmollah\" target=\"_blank\">@abdurrakibmollah</a> </p>",
          "rawMarkdown": "Welcome @abdurrakibmollah "
        }
      ]
    },
    {
      "id": 1805470,
      "postDate": "2022-05-30T08:07:06.283Z",
      "content": "<p>Thanks for sharing!</p>",
      "rawMarkdown": "Thanks for sharing!",
      "votes": 1,
      "replies": [
        {
          "id": 1805487,
          "postDate": "2022-05-30T08:28:58.777Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/anuragmallick\" target=\"_blank\">@anuragmallick</a> </p>",
          "rawMarkdown": "Welcome @anuragmallick "
        }
      ]
    },
    {
      "id": 1805076,
      "postDate": "2022-05-29T18:57:08.093Z",
      "content": "<p>Thanks for sharing!it helps a lot</p>",
      "rawMarkdown": "Thanks for sharing!it helps a lot\n",
      "votes": 1,
      "replies": [
        {
          "id": 1805488,
          "postDate": "2022-05-30T08:29:12.137Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/maleklaatiri\" target=\"_blank\">@maleklaatiri</a> </p>",
          "rawMarkdown": "Welcome @maleklaatiri "
        }
      ]
    },
    {
      "id": 1804363,
      "postDate": "2022-05-28T20:12:36.343Z",
      "content": "<p>Thank you for sharing this with us! Very helpful!</p>",
      "rawMarkdown": "Thank you for sharing this with us! Very helpful!",
      "votes": 1,
      "replies": [
        {
          "id": 1804390,
          "postDate": "2022-05-28T21:46:21.960Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/datark1\" target=\"_blank\">@datark1</a> </p>",
          "rawMarkdown": "Welcome @datark1 "
        }
      ]
    },
    {
      "id": 1804333,
      "postDate": "2022-05-28T19:17:22.933Z",
      "content": "<p>Thanks for sharing!</p>",
      "rawMarkdown": "Thanks for sharing!",
      "votes": 1,
      "replies": [
        {
          "id": 1804347,
          "postDate": "2022-05-28T19:36:05.600Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/amishanand\" target=\"_blank\">@amishanand</a> </p>",
          "rawMarkdown": "Welcome @amishanand "
        }
      ]
    },
    {
      "id": 1803951,
      "postDate": "2022-05-28T11:25:38.970Z",
      "content": "<p>thanks for sharing this! <br>\nIt will be perfect with <code>train_labels.csv</code> added.</p>",
      "rawMarkdown": "thanks for sharing this! \nIt will be perfect with `train_labels.csv` added.",
      "votes": 1,
      "replies": [
        {
          "id": 1804172,
          "postDate": "2022-05-28T15:50:22.793Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/cherrizhu\" target=\"_blank\">@cherrizhu</a> <br>\nI have already merged the labels in the <code>train_data.parquet</code> file </p>",
          "rawMarkdown": "Welcome @cherrizhu \nI have already merged the labels in the `train_data.parquet` file ",
          "votes": 1
        }
      ]
    },
    {
      "id": 1803947,
      "postDate": "2022-05-28T11:11:09.820Z",
      "content": "<p>Thanks for sharing!</p>",
      "rawMarkdown": "Thanks for sharing!",
      "votes": 1,
      "replies": [
        {
          "id": 1804173,
          "postDate": "2022-05-28T15:50:36.903Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/yuezhoul\" target=\"_blank\">@yuezhoul</a> </p>",
          "rawMarkdown": "Welcome @yuezhoul "
        }
      ]
    },
    {
      "id": 1803856,
      "postDate": "2022-05-28T08:49:47.120Z",
      "content": "<p>Thank you for sharing this!</p>",
      "rawMarkdown": "Thank you for sharing this!",
      "votes": 1,
      "replies": [
        {
          "id": 1803943,
          "postDate": "2022-05-28T11:02:14.463Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/rajaahdjey\" target=\"_blank\">@rajaahdjey</a> </p>",
          "rawMarkdown": "Welcome @rajaahdjey "
        }
      ]
    },
    {
      "id": 1803528,
      "postDate": "2022-05-27T23:10:15.173Z",
      "content": "<p>Thanks for the datasets 👍</p>",
      "rawMarkdown": "Thanks for the datasets 👍",
      "votes": 1,
      "replies": [
        {
          "id": 1803548,
          "postDate": "2022-05-27T23:57:52.600Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/jeanmidev\" target=\"_blank\">@jeanmidev</a> </p>",
          "rawMarkdown": "Welcome @jeanmidev "
        }
      ]
    },
    {
      "id": 1803081,
      "postDate": "2022-05-27T12:46:36.100Z",
      "content": "<p>Thanks for sharing</p>",
      "rawMarkdown": "Thanks for sharing",
      "votes": 1,
      "replies": [
        {
          "id": 1803096,
          "postDate": "2022-05-27T13:17:36.350Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/muhhizbulzainulm\" target=\"_blank\">@muhhizbulzainulm</a> </p>",
          "rawMarkdown": "Welcome @muhhizbulzainulm "
        }
      ]
    },
    {
      "id": 1802982,
      "postDate": "2022-05-27T10:57:43.627Z",
      "content": "<p>I understand that you reduced the size of the file by converting it to parquet format and using float32.<br>\nThat's a great learning experience! Thank you!🙌</p>",
      "rawMarkdown": "I understand that you reduced the size of the file by converting it to parquet format and using float32.\nThat's a great learning experience! Thank you!🙌",
      "votes": 1,
      "replies": [
        {
          "id": 1803008,
          "postDate": "2022-05-27T11:14:07.850Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/tgwstr\" target=\"_blank\">@tgwstr</a> </p>",
          "rawMarkdown": "Welcome @tgwstr "
        }
      ]
    },
    {
      "id": 1802248,
      "postDate": "2022-05-26T15:32:17.583Z",
      "content": "<p>Thank you!! It would be very helpful :)</p>",
      "rawMarkdown": "Thank you!! It would be very helpful :)",
      "votes": 1,
      "replies": [
        {
          "id": 1802255,
          "postDate": "2022-05-26T15:41:25.830Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/sangwooooo\" target=\"_blank\">@sangwooooo</a> </p>",
          "rawMarkdown": "Welcome @sangwooooo "
        }
      ]
    },
    {
      "id": 1802186,
      "postDate": "2022-05-26T14:36:28.530Z",
      "content": "<p>Thanks for sharing it!👍.If possible can you share some resources regarding this type of file conversion.I really want to learn.😄  </p>",
      "rawMarkdown": "Thanks for sharing it!👍.If possible can you share some resources regarding this type of file conversion.I really want to learn.😄  ",
      "votes": 1,
      "replies": [
        {
          "id": 1802224,
          "postDate": "2022-05-26T14:56:08.203Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/gomohit\" target=\"_blank\">@gomohit</a> </p>",
          "rawMarkdown": "Welcome @gomohit "
        }
      ]
    },
    {
      "id": 1802099,
      "postDate": "2022-05-26T12:57:52.683Z",
      "content": "<p>Thank you for sharing</p>",
      "rawMarkdown": "Thank you for sharing",
      "votes": 1,
      "replies": [
        {
          "id": 1802133,
          "postDate": "2022-05-26T13:33:13.440Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/thedevastator\" target=\"_blank\">@thedevastator</a> </p>",
          "rawMarkdown": "Welcome @thedevastator "
        }
      ]
    },
    {
      "id": 1802048,
      "postDate": "2022-05-26T11:46:01.223Z",
      "content": "<p>thank you very much 👍</p>",
      "rawMarkdown": "thank you very much 👍",
      "votes": 1,
      "replies": [
        {
          "id": 1802132,
          "postDate": "2022-05-26T13:33:03.817Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/sanctuarywu\" target=\"_blank\">@sanctuarywu</a> </p>",
          "rawMarkdown": "Welcome @sanctuarywu "
        }
      ]
    },
    {
      "id": 1801716,
      "postDate": "2022-05-26T04:47:00.413Z",
      "content": "<p>Thanks for sharing it</p>",
      "rawMarkdown": "Thanks for sharing it",
      "votes": 1,
      "replies": [
        {
          "id": 1801749,
          "postDate": "2022-05-26T05:39:50.733Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/devkhant24\" target=\"_blank\">@devkhant24</a> </p>",
          "rawMarkdown": "Welcome @devkhant24 ",
          "votes": 1
        }
      ]
    },
    {
      "id": 1801688,
      "postDate": "2022-05-26T03:34:59.267Z",
      "content": "<p>Thanks for this, makes life a lot easier 🙌</p>",
      "rawMarkdown": "Thanks for this, makes life a lot easier 🙌",
      "votes": 1,
      "replies": [
        {
          "id": 1801751,
          "postDate": "2022-05-26T05:39:59.167Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/niekvanderzwaag\" target=\"_blank\">@niekvanderzwaag</a> </p>",
          "rawMarkdown": "Welcome @niekvanderzwaag ",
          "votes": 1
        }
      ]
    },
    {
      "id": 1838452,
      "postDate": "2022-06-30T14:38:48.433Z",
      "content": "<p>will parquet link be considered as external dataset. submission code should be on real data. isn't is the case?</p>",
      "rawMarkdown": "will parquet link be considered as external dataset. submission code should be on real data. isn't is the case?",
      "votes": 2,
      "replies": [
        {
          "id": 1838745,
          "postDate": "2022-06-30T19:09:21.023Z",
          "content": "<p>Hi <a href=\"https://www.kaggle.com/kartikasharma\" target=\"_blank\">@kartikasharma</a> , Publicly available external data is allowed in this competition </p>",
          "rawMarkdown": "Hi @kartikasharma , Publicly available external data is allowed in this competition ",
          "votes": 2
        }
      ]
    },
    {
      "id": 1820263,
      "postDate": "2022-06-14T13:25:14.053Z",
      "content": "<p>Thanks for sharing the datasets!!</p>",
      "rawMarkdown": "Thanks for sharing the datasets!!",
      "votes": 2,
      "replies": [
        {
          "id": 1820270,
          "postDate": "2022-06-14T13:34:32.110Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/nijianzhang\" target=\"_blank\">@nijianzhang</a> </p>",
          "rawMarkdown": "Welcome @nijianzhang ",
          "votes": 1
        }
      ]
    },
    {
      "id": 1816844,
      "postDate": "2022-06-10T15:58:22.200Z",
      "content": "<p>Thanks for sharing</p>",
      "rawMarkdown": "Thanks for sharing",
      "votes": 2,
      "replies": [
        {
          "id": 1816980,
          "postDate": "2022-06-10T18:16:32.877Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/medgic\" target=\"_blank\">@medgic</a> </p>",
          "rawMarkdown": "Welcome @medgic ",
          "votes": 1
        }
      ]
    },
    {
      "id": 1808544,
      "postDate": "2022-06-02T01:14:22.090Z",
      "content": "<p>Thank you for sharing the datasets</p>",
      "rawMarkdown": "Thank you for sharing the datasets",
      "votes": 2,
      "replies": [
        {
          "id": 1809682,
          "postDate": "2022-06-03T00:11:43.483Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/vishnukonjeti\" target=\"_blank\">@vishnukonjeti</a> </p>",
          "rawMarkdown": "Welcome @vishnukonjeti ",
          "votes": 1
        }
      ]
    },
    {
      "id": 1802603,
      "postDate": "2022-05-27T00:10:09.313Z",
      "content": "<p>Thank you! It helps a lot!</p>",
      "rawMarkdown": "Thank you! It helps a lot!",
      "votes": 2,
      "replies": [
        {
          "id": 1802855,
          "postDate": "2022-05-27T08:05:32.537Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/arti1117\" target=\"_blank\">@arti1117</a> </p>",
          "rawMarkdown": "Welcome @arti1117 "
        }
      ]
    },
    {
      "id": 1802187,
      "postDate": "2022-05-26T14:38:01.593Z",
      "content": "<p>Thanks very much for sharing. Great to see the memory improvements. Would you recommend converting the categorical features as well?</p>",
      "rawMarkdown": "Thanks very much for sharing. Great to see the memory improvements. Would you recommend converting the categorical features as well?",
      "votes": 2,
      "replies": [
        {
          "id": 1802222,
          "postDate": "2022-05-26T14:55:55.220Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/datajmcn\" target=\"_blank\">@datajmcn</a> </p>",
          "rawMarkdown": "Welcome @datajmcn ",
          "votes": 1
        }
      ]
    },
    {
      "id": 1801775,
      "postDate": "2022-05-26T06:24:00.180Z",
      "content": "<p>Hi Sanskar, this might be a lazy comment as I did not compare both the datasets, but have you checked if the values before and after conversion are same and there aren't any overflow issues ? Anyways thanks a lot, will save my life</p>",
      "rawMarkdown": "Hi Sanskar, this might be a lazy comment as I did not compare both the datasets, but have you checked if the values before and after conversion are same and there aren't any overflow issues ? Anyways thanks a lot, will save my life",
      "votes": 2,
      "replies": [
        {
          "id": 1801913,
          "postDate": "2022-05-26T09:10:48.740Z",
          "content": "<p>Thanks, <a href=\"https://www.kaggle.com/nikhilmishradev\" target=\"_blank\">@nikhilmishradev</a> . I have not compared both the file formats. I will check and provide an update here </p>",
          "rawMarkdown": "Thanks, @nikhilmishradev . I have not compared both the file formats. I will check and provide an update here ",
          "votes": 1
        },
        {
          "id": 1802278,
          "postDate": "2022-05-26T16:00:31.523Z",
          "content": "<p>Hi <a href=\"https://www.kaggle.com/nikhilmishradev\" target=\"_blank\">@nikhilmishradev</a>, the value after data format conversion changes a bit( gets rounded off ). This is mainly because of the conversion of <code>float64</code> to <code>float32</code> values. <br>\nCheck the following image for reference: <br>\n<img src=\"https://raw.githubusercontent.com/sanskar-hasija/kaggle/main/images/amex_data_conversion.png\" alt=\"\"></p>",
          "rawMarkdown": "Hi @nikhilmishradev, the value after data format conversion changes a bit( gets rounded off ). This is mainly because of the conversion of `float64` to `float32` values. \nCheck the following image for reference: \n![](https://raw.githubusercontent.com/sanskar-hasija/kaggle/main/images/amex_data_conversion.png)",
          "votes": 5
        }
      ]
    },
    {
      "id": 1801487,
      "postDate": "2022-05-25T19:46:39.440Z",
      "content": "<p>Now ,Maybe it won't gonna show us you are running out of memory.Thanks,for saving kagglers as well as their system.</p>",
      "rawMarkdown": "Now ,Maybe it won't gonna show us you are running out of memory.Thanks,for saving kagglers as well as their system.",
      "votes": 2,
      "replies": [
        {
          "id": 1801490,
          "postDate": "2022-05-25T19:53:54.660Z",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/ramjasmaurya\" target=\"_blank\">@ramjasmaurya</a> </p>",
          "rawMarkdown": "Welcome @ramjasmaurya ",
          "votes": 1
        }
      ]
    },
    {
      "id": 2044765,
      "postDate": "2022-11-26T18:48:42.413Z",
      "content": "<p>Thanks for sharing, looking forward to explore this option. </p>",
      "rawMarkdown": "Thanks for sharing, looking forward to explore this option. "
    },
    {
      "id": 2023201,
      "postDate": "2022-11-09T15:53:17.913Z",
      "rawMarkdown": "",
      "isDeleted": true
    },
    {
      "id": 2627235,
      "postDate": "2024-01-30T14:50:29.357Z",
      "content": "<p>Thank you for sharing.</p>",
      "rawMarkdown": "Thank you for sharing.",
      "votes": 1
    },
    {
      "id": 1906072,
      "postDate": "2022-08-19T14:49:11.720Z",
      "content": "<p>Thanks for sharing <a href=\"https://www.kaggle.com/odins0n\" target=\"_blank\">@odins0n</a>. </p>",
      "rawMarkdown": "Thanks for sharing @odins0n. ",
      "votes": 1
    },
    {
      "id": 1837986,
      "postDate": "2022-06-30T05:14:34.887Z",
      "content": "<p>It's very useful! <br>\nThanks!</p>",
      "rawMarkdown": "It's very useful! \nThanks!",
      "votes": 1
    },
    {
      "id": 1821585,
      "postDate": "2022-06-15T17:47:40.233Z",
      "content": "<p>Thanks for sharing <a href=\"https://www.kaggle.com/odins0n\" target=\"_blank\">@odins0n</a> </p>",
      "rawMarkdown": "Thanks for sharing @odins0n ",
      "votes": 1
    }
  ],
  "comments": [
    {
      "id": 1843525,
      "author_name": "The Devastator",
      "author_url": "",
      "post_date": "2022-07-05T01:45:47.093000",
      "content": "<p><img src=\"https://i.ibb.co/3hHnF5f/Untitled.png\" alt=\"\"></p>",
      "votes": 4,
      "replies": []
    },
    {
      "id": 1897923,
      "author_name": "cam",
      "author_url": "",
      "post_date": "2022-08-14T06:43:33.310000",
      "content": "<p>that's very useful!</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 1833179,
      "author_name": "Nith_in",
      "author_url": "",
      "post_date": "2022-06-25T17:48:06.860000",
      "content": "<p>Thanks for sharing <a href=\"https://www.kaggle.com/odins0n\" target=\"_blank\">@odins0n</a>. Could you please tell how it was converted to parquet file?</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 1807181,
      "author_name": "MIkhail Donskoy",
      "author_url": "",
      "post_date": "2022-05-31T19:46:25.737000",
      "content": "<p>Thank you!! It would be very helpful!</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1809677,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-06-03T00:08:56.447000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/michaelpoluektov\" target=\"_blank\">@michaelpoluektov</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1806783,
      "author_name": "Michael Poluektov",
      "author_url": "",
      "post_date": "2022-05-31T13:11:35.480000",
      "content": "<p>Thanks for sharing! I noticed that <a href=\"https://www.kaggle.com/ruchi798\" target=\"_blank\">@ruchi798</a> has suggested using feather files with float16 <a href=\"https://www.kaggle.com/competitions/amex-default-prediction/discussion/327143\" target=\"_blank\">here</a>. I was wondering, since the values have been normalised and random noise has supposedly been added, do you think there are any advantages of using float32 instead of float16?</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1809681,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-06-03T00:11:29.777000",
          "content": "<p>It will depend from feature to feature but in general yes. Float32 will perform better than float16 since more and more information will be lost when we compress it from the original float64 values. </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1806446,
      "author_name": "IMvision12",
      "author_url": "",
      "post_date": "2022-05-31T07:13:22.830000",
      "content": "<p>Thanks for sharing it</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1809676,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-06-03T00:08:43.110000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/imvision12\" target=\"_blank\">@imvision12</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1806288,
      "author_name": "Pedro Souza",
      "author_url": "",
      "post_date": "2022-05-31T02:38:12.457000",
      "content": "<p>Thank you for sharing</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1806290,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-31T02:42:46.403000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/pedrohsouzax\" target=\"_blank\">@pedrohsouzax</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1806258,
      "author_name": "mungeunjo",
      "author_url": "",
      "post_date": "2022-05-31T01:18:30.747000",
      "content": "<p>Thanks for sharing brother</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1806259,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-31T01:22:44.500000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/mungeunjo\" target=\"_blank\">@mungeunjo</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1806120,
      "author_name": "Abdur Rakib Mollah",
      "author_url": "",
      "post_date": "2022-05-30T19:57:11.817000",
      "content": "<p>Thanks for this.</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1806260,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-31T01:22:51.340000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/abdurrakibmollah\" target=\"_blank\">@abdurrakibmollah</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1805470,
      "author_name": "ANURAG MALLICK",
      "author_url": "",
      "post_date": "2022-05-30T08:07:06.283000",
      "content": "<p>Thanks for sharing!</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1805487,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-30T08:28:58.777000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/anuragmallick\" target=\"_blank\">@anuragmallick</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1805076,
      "author_name": "Vinci",
      "author_url": "",
      "post_date": "2022-05-29T18:57:08.093000",
      "content": "<p>Thanks for sharing!it helps a lot</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1805488,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-30T08:29:12.137000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/maleklaatiri\" target=\"_blank\">@maleklaatiri</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1804363,
      "author_name": "Robert Kwiatkowski",
      "author_url": "",
      "post_date": "2022-05-28T20:12:36.343000",
      "content": "<p>Thank you for sharing this with us! Very helpful!</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1804390,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-28T21:46:21.960000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/datark1\" target=\"_blank\">@datark1</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1804333,
      "author_name": "Amish Anand",
      "author_url": "",
      "post_date": "2022-05-28T19:17:22.933000",
      "content": "<p>Thanks for sharing!</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1804347,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-28T19:36:05.600000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/amishanand\" target=\"_blank\">@amishanand</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1803951,
      "author_name": "chuan",
      "author_url": "",
      "post_date": "2022-05-28T11:25:38.970000",
      "content": "<p>thanks for sharing this! <br>\nIt will be perfect with <code>train_labels.csv</code> added.</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1804172,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-28T15:50:22.793000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/cherrizhu\" target=\"_blank\">@cherrizhu</a> <br>\nI have already merged the labels in the <code>train_data.parquet</code> file </p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 1803947,
      "author_name": "yuezhouL",
      "author_url": "",
      "post_date": "2022-05-28T11:11:09.820000",
      "content": "<p>Thanks for sharing!</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1804173,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-28T15:50:36.903000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/yuezhoul\" target=\"_blank\">@yuezhoul</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1803856,
      "author_name": "Rajaah Djey",
      "author_url": "",
      "post_date": "2022-05-28T08:49:47.120000",
      "content": "<p>Thank you for sharing this!</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1803943,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-28T11:02:14.463000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/rajaahdjey\" target=\"_blank\">@rajaahdjey</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1803528,
      "author_name": "Jean-Michel D.",
      "author_url": "",
      "post_date": "2022-05-27T23:10:15.173000",
      "content": "<p>Thanks for the datasets 👍</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1803548,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-27T23:57:52.600000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/jeanmidev\" target=\"_blank\">@jeanmidev</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1803081,
      "author_name": "Muh. Hizbul Zainul M",
      "author_url": "",
      "post_date": "2022-05-27T12:46:36.100000",
      "content": "<p>Thanks for sharing</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1803096,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-27T13:17:36.350000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/muhhizbulzainulm\" target=\"_blank\">@muhhizbulzainulm</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1802982,
      "author_name": "chinchilla",
      "author_url": "",
      "post_date": "2022-05-27T10:57:43.627000",
      "content": "<p>I understand that you reduced the size of the file by converting it to parquet format and using float32.<br>\nThat's a great learning experience! Thank you!🙌</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1803008,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-27T11:14:07.850000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/tgwstr\" target=\"_blank\">@tgwstr</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1802248,
      "author_name": "Llama_4",
      "author_url": "",
      "post_date": "2022-05-26T15:32:17.583000",
      "content": "<p>Thank you!! It would be very helpful :)</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1802255,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-26T15:41:25.830000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/sangwooooo\" target=\"_blank\">@sangwooooo</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1802186,
      "author_name": "Mohit",
      "author_url": "",
      "post_date": "2022-05-26T14:36:28.530000",
      "content": "<p>Thanks for sharing it!👍.If possible can you share some resources regarding this type of file conversion.I really want to learn.😄  </p>",
      "votes": 1,
      "replies": [
        {
          "id": 1802224,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-26T14:56:08.203000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/gomohit\" target=\"_blank\">@gomohit</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1802099,
      "author_name": "The Devastator",
      "author_url": "",
      "post_date": "2022-05-26T12:57:52.683000",
      "content": "<p>Thank you for sharing</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1802133,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-26T13:33:13.440000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/thedevastator\" target=\"_blank\">@thedevastator</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1802048,
      "author_name": "sanctuary_wu",
      "author_url": "",
      "post_date": "2022-05-26T11:46:01.223000",
      "content": "<p>thank you very much 👍</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1802132,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-26T13:33:03.817000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/sanctuarywu\" target=\"_blank\">@sanctuarywu</a> </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1801716,
      "author_name": "Dev Khant",
      "author_url": "",
      "post_date": "2022-05-26T04:47:00.413000",
      "content": "<p>Thanks for sharing it</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1801749,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-26T05:39:50.733000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/devkhant24\" target=\"_blank\">@devkhant24</a> </p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 1801688,
      "author_name": "Niek van der Zwaag",
      "author_url": "",
      "post_date": "2022-05-26T03:34:59.267000",
      "content": "<p>Thanks for this, makes life a lot easier 🙌</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1801751,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-05-26T05:39:59.167000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/niekvanderzwaag\" target=\"_blank\">@niekvanderzwaag</a> </p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 1838452,
      "author_name": "kartika",
      "author_url": "",
      "post_date": "2022-06-30T14:38:48.433000",
      "content": "<p>will parquet link be considered as external dataset. submission code should be on real data. isn't is the case?</p>",
      "votes": 2,
      "replies": [
        {
          "id": 1838745,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-06-30T19:09:21.023000",
          "content": "<p>Hi <a href=\"https://www.kaggle.com/kartikasharma\" target=\"_blank\">@kartikasharma</a> , Publicly available external data is allowed in this competition </p>",
          "votes": 2,
          "replies": []
        }
      ]
    },
    {
      "id": 1820263,
      "author_name": "Jianzhang Ni",
      "author_url": "",
      "post_date": "2022-06-14T13:25:14.053000",
      "content": "<p>Thanks for sharing the datasets!!</p>",
      "votes": 2,
      "replies": [
        {
          "id": 1820270,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-06-14T13:34:32.110000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/nijianzhang\" target=\"_blank\">@nijianzhang</a> </p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 1816844,
      "author_name": "ÇiğdemRenkli",
      "author_url": "",
      "post_date": "2022-06-10T15:58:22.200000",
      "content": "<p>Thanks for sharing</p>",
      "votes": 2,
      "replies": [
        {
          "id": 1816980,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-06-10T18:16:32.877000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/medgic\" target=\"_blank\">@medgic</a> </p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 1808544,
      "author_name": "Vishnu Konjeti",
      "author_url": "",
      "post_date": "2022-06-02T01:14:22.090000",
      "content": "<p>Thank you for sharing the datasets</p>",
      "votes": 2,
      "replies": [
        {
          "id": 1809682,
          "author_name": "Sanskar Hasija",
          "author_url": "",
          "post_date": "2022-06-03T00:11:43.483000",
          "content": "<p>Welcome <a href=\"https://www.kaggle.com/vishnukonjeti\" target=\"_blank\">@vishnukonjeti</a> </p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 1802603,
      "author_name": "",
      "author_url": "",
      "post_date": "2022-05-27T00:10:09.313000",
      "content": "",
      "votes": 2,
      "replies": [
        {
          "id": 1802855,
          "author_name": "",
          "author_url": "",
          "post_date": "2022-05-27T08:05:32.537000",
          "content": "",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1802187,
      "author_name": "",
      "author_url": "",
      "post_date": "2022-05-26T14:38:01.593000",
      "content": "",
      "votes": 2,
      "replies": [
        {
          "id": 1802222,
          "author_name": "",
          "author_url": "",
          "post_date": "2022-05-26T14:55:55.220000",
          "content": "",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 1801775,
      "author_name": "",
      "author_url": "",
      "post_date": "2022-05-26T06:24:00.180000",
      "content": "",
      "votes": 2,
      "replies": [
        {
          "id": 1801913,
          "author_name": "",
          "author_url": "",
          "post_date": "2022-05-26T09:10:48.740000",
          "content": "",
          "votes": 1,
          "replies": []
        },
        {
          "id": 1802278,
          "author_name": "",
          "author_url": "",
          "post_date": "2022-05-26T16:00:31.523000",
          "content": "",
          "votes": 5,
          "replies": []
        }
      ]
    },
    {
      "id": 1801487,
      "author_name": "",
      "author_url": "",
      "post_date": "2022-05-25T19:46:39.440000",
      "content": "",
      "votes": 2,
      "replies": [
        {
          "id": 1801490,
          "author_name": "",
          "author_url": "",
          "post_date": "2022-05-25T19:53:54.660000",
          "content": "",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 2044765,
      "author_name": "",
      "author_url": "",
      "post_date": "2022-11-26T18:48:42.413000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 2023201,
      "author_name": "",
      "author_url": "",
      "post_date": "2022-11-09T15:53:17.913000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 2627235,
      "author_name": "",
      "author_url": "",
      "post_date": "2024-01-30T14:50:29.357000",
      "content": "",
      "votes": 1,
      "replies": []
    },
    {
      "id": 1906072,
      "author_name": "",
      "author_url": "",
      "post_date": "2022-08-19T14:49:11.720000",
      "content": "",
      "votes": 1,
      "replies": []
    },
    {
      "id": 1837986,
      "author_name": "",
      "author_url": "",
      "post_date": "2022-06-30T05:14:34.887000",
      "content": "",
      "votes": 1,
      "replies": []
    },
    {
      "id": 1821585,
      "author_name": "",
      "author_url": "",
      "post_date": "2022-06-15T17:47:40.233000",
      "content": "",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1801471": "Hello, I have created Parquet format files for train_data.csv and test_data.csv files. The size of the CSV files decreased as follows:\nFull Dataset: <b>50.31 GB</b> ➡️ <b>10.41 GB</b>\ntrain_data : <b>16.39 GB</b> ➡️ <b>3.35 GB</b>\ntest_data: <b>33.82 GB</b> ➡️ <b>6.96 GB</b>\n\n\nLink to Dataset - www.kaggle.com/odins0n/amex-parquet\n\nThe columns with <b>float64</b> dtype were also converted to <b>float32</b> dtype as well.  Parquet files are much faster to read in for datasets of this size\n\nExample Notebook for loading Parquet files - www.kaggle.com/code/odins0n/load-parquet-files-with-low-memory/\nPlease let me know if you have any feedback 😀\n",
    "1843525": "![](https://i.ibb.co/3hHnF5f/Untitled.png)",
    "1897923": "that's very useful!",
    "1833179": "Thanks for sharing @odins0n. Could you please tell how it was converted to parquet file?",
    "1807181": "Thank you!! It would be very helpful!",
    "1806783": "Thanks for sharing! I noticed that @ruchi798 has suggested using feather files with float16 [here](https://www.kaggle.com/competitions/amex-default-prediction/discussion/327143). I was wondering, since the values have been normalised and random noise has supposedly been added, do you think there are any advantages of using float32 instead of float16?",
    "1806446": "Thanks for sharing it",
    "1806288": "Thank you for sharing",
    "1806258": "Thanks for sharing brother",
    "1806120": "Thanks for this.",
    "1805470": "Thanks for sharing!",
    "1805076": "Thanks for sharing!it helps a lot\n",
    "1804363": "Thank you for sharing this with us! Very helpful!",
    "1804333": "Thanks for sharing!",
    "1803951": "thanks for sharing this! \nIt will be perfect with `train_labels.csv` added.",
    "1803947": "Thanks for sharing!",
    "1803856": "Thank you for sharing this!",
    "1803528": "Thanks for the datasets 👍",
    "1803081": "Thanks for sharing",
    "1802982": "I understand that you reduced the size of the file by converting it to parquet format and using float32.\nThat's a great learning experience! Thank you!🙌",
    "1802248": "Thank you!! It would be very helpful :)",
    "1802186": "Thanks for sharing it!👍.If possible can you share some resources regarding this type of file conversion.I really want to learn.😄  ",
    "1802099": "Thank you for sharing",
    "1802048": "thank you very much 👍",
    "1801716": "Thanks for sharing it",
    "1801688": "Thanks for this, makes life a lot easier 🙌",
    "1838452": "will parquet link be considered as external dataset. submission code should be on real data. isn't is the case?",
    "1820263": "Thanks for sharing the datasets!!",
    "1816844": "Thanks for sharing",
    "1808544": "Thank you for sharing the datasets",
    "1802603": "Thank you! It helps a lot!",
    "1802187": "Thanks very much for sharing. Great to see the memory improvements. Would you recommend converting the categorical features as well?",
    "1801775": "Hi Sanskar, this might be a lazy comment as I did not compare both the datasets, but have you checked if the values before and after conversion are same and there aren't any overflow issues ? Anyways thanks a lot, will save my life",
    "1801487": "Now ,Maybe it won't gonna show us you are running out of memory.Thanks,for saving kagglers as well as their system.",
    "2044765": "Thanks for sharing, looking forward to explore this option. ",
    "2023201": "",
    "2627235": "Thank you for sharing.",
    "1906072": "Thanks for sharing @odins0n. ",
    "1837986": "It's very useful! \nThanks!",
    "1821585": "Thanks for sharing @odins0n "
  }
}