{
  "id": 386180,
  "title": "Dataset for conducting EDA without memory constraints of 8GB 👍",
  "url": "/competitions/predict-student-performance-from-game-play/discussion/386180",
  "author_name": "",
  "post_date": "2023-02-11T16:03:55.201956100Z",
  "votes": 6,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I share <a href=\"https://www.kaggle.com/datasets/paulbacher/pred-student-perf-without-competiton-env\" target=\"_blank\">here</a> the competition dataset <strong>without</strong> the <code>jo_wilder</code> folder so you can use the 30GB of the CPU. It may be usefull for EDA or other debugging.<br>\nGood luck</p>",
  "messages": [
    {
      "id": "2140289",
      "postDate": "02/11/2023 16:03:55",
      "content": "<p>I share <a href=\"https://www.kaggle.com/datasets/paulbacher/pred-student-perf-without-competiton-env\" target=\"_blank\">here</a> the competition dataset <strong>without</strong> the <code>jo_wilder</code> folder so you can use the 30GB of the CPU. It may be usefull for EDA or other debugging.<br>\nGood luck</p>",
      "rawMarkdown": "I share [here](https://www.kaggle.com/datasets/paulbacher/pred-student-perf-without-competiton-env) the competition dataset **without** the `jo_wilder` folder so you can use the 30GB of the CPU. It may be usefull for EDA or other debugging.\nGood luck",
      "votes": null
    },
    {
      "id": "2140401",
      "postDate": "02/11/2023 17:51:48",
      "content": "<p>Thanks. Another solution is to use a GPU notebook with the <code>jo_wilder</code> folder, then the VM will have 16GB VRAM and 32GB RAM for total of 48GB RAM. And it won't be reduced to 8GB RAM.</p>\n<p>Furthermore, we can use GPU to feature engineer and train a model with GPU. Then we save the model. Then we load the model into a second CPU inference notebook. </p>",
      "rawMarkdown": "Thanks. Another solution is to use a GPU notebook with the `jo_wilder` folder, then the VM will have 16GB VRAM and 32GB RAM for total of 48GB RAM. And it won't be reduced to 8GB RAM.\n\nFurthermore, we can use GPU to feature engineer and train a model with GPU. Then we save the model. Then we load the model into a second CPU inference notebook.",
      "votes": null
    },
    {
      "id": "2140914",
      "postDate": "02/12/2023 09:21:41",
      "content": "<p>Thanks Chris for this alternative solution. I'm running out of GPU credits this week but I'll think about that later.</p>",
      "rawMarkdown": "Thanks Chris for this alternative solution. I'm running out of GPU credits this week but I'll think about that later.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2140401,
      "author_name": "cdeotte",
      "author_url": "",
      "post_date": "02/11/2023 17:51:48",
      "content": "<p>Thanks. Another solution is to use a GPU notebook with the <code>jo_wilder</code> folder, then the VM will have 16GB VRAM and 32GB RAM for total of 48GB RAM. And it won't be reduced to 8GB RAM.</p>\n<p>Furthermore, we can use GPU to feature engineer and train a model with GPU. Then we save the model. Then we load the model into a second CPU inference notebook. </p>",
      "votes": null,
      "replies": [
        {
          "id": 2140914,
          "author_name": "paulbacher",
          "author_url": "",
          "post_date": "02/12/2023 09:21:41",
          "content": "<p>Thanks Chris for this alternative solution. I'm running out of GPU credits this week but I'll think about that later.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2140289": "I share [here](https://www.kaggle.com/datasets/paulbacher/pred-student-perf-without-competiton-env) the competition dataset **without** the `jo_wilder` folder so you can use the 30GB of the CPU. It may be usefull for EDA or other debugging.\nGood luck",
    "2140401": "Thanks. Another solution is to use a GPU notebook with the `jo_wilder` folder, then the VM will have 16GB VRAM and 32GB RAM for total of 48GB RAM. And it won't be reduced to 8GB RAM.\n\nFurthermore, we can use GPU to feature engineer and train a model with GPU. Then we save the model. Then we load the model into a second CPU inference notebook.",
    "2140914": "Thanks Chris for this alternative solution. I'm running out of GPU credits this week but I'll think about that later."
  },
  "source": "meta"
}