{
  "id": 507894,
  "title": "keep getting oom error",
  "url": "/competitions/home-credit-credit-risk-model-stability/discussion/507894",
  "author_name": "",
  "post_date": "2024-05-27T17:24:45.703024Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hey guys, first-timer here. I was just wondering why OOM error always appear as I had del almost all objects and dataframes during training process. And how large is the hidden test set. <br>\nNow I ran out of my submissions…</p>",
  "messages": [
    {
      "id": "2839763",
      "postDate": "05/27/2024 17:24:45",
      "content": "<p>Hey guys, first-timer here. I was just wondering why OOM error always appear as I had del almost all objects and dataframes during training process. And how large is the hidden test set. <br>\nNow I ran out of my submissions…</p>",
      "rawMarkdown": "Hey guys, first-timer here. I was just wondering why OOM error always appear as I had del almost all objects and dataframes during training process. And how large is the hidden test set. \nNow I ran out of my submissions...",
      "votes": null
    },
    {
      "id": "2839887",
      "postDate": "05/27/2024 18:44:57",
      "content": "<p>It's better to store trained model and load them rather than submitting with training. Also it's better to predict in batches if you have large model or model sets.</p>",
      "rawMarkdown": "It's better to store trained model and load them rather than submitting with training. Also it's better to predict in batches if you have large model or model sets.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2839887,
      "author_name": "shyhandsome",
      "author_url": "",
      "post_date": "05/27/2024 18:44:57",
      "content": "<p>It's better to store trained model and load them rather than submitting with training. Also it's better to predict in batches if you have large model or model sets.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2839763": "Hey guys, first-timer here. I was just wondering why OOM error always appear as I had del almost all objects and dataframes during training process. And how large is the hidden test set. \nNow I ran out of my submissions...",
    "2839887": "It's better to store trained model and load them rather than submitting with training. Also it's better to predict in batches if you have large model or model sets."
  },
  "source": "meta"
}