{
  "id": 133612,
  "title": "How many models can be run in the submission kernel?",
  "url": "/competitions/bengaliai-cv19/discussion/133612",
  "author_name": "",
  "post_date": "2020-03-03T13:49:23.021520300Z",
  "votes": 1,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I try to run inferences of multiple models to ensemble the outputs in a submission kernel.\nBut it is hard to run long, or heavy inferences cause it may occur submission errors (i.e. Notebook Exceeded Allowed Compute).\nDid anyone check the maximum number of models possible in one-time kernel?</p>",
  "messages": [
    {
      "id": "762410",
      "postDate": "03/03/2020 13:49:23",
      "content": "<p>I try to run inferences of multiple models to ensemble the outputs in a submission kernel.\nBut it is hard to run long, or heavy inferences cause it may occur submission errors (i.e. Notebook Exceeded Allowed Compute).\nDid anyone check the maximum number of models possible in one-time kernel?</p>",
      "rawMarkdown": "I try to run inferences of multiple models to ensemble the outputs in a submission kernel.\nBut it is hard to run long, or heavy inferences cause it may occur submission errors (i.e. Notebook Exceeded Allowed Compute).\nDid anyone check the maximum number of models possible in one-time kernel?",
      "votes": null
    },
    {
      "id": "762452",
      "postDate": "03/03/2020 14:20:50",
      "content": "<p>It depends. I think 5-10 models/folds at least, if you optimize your inference kernel, probably you can ensemble more.\nTake a look at <a href=\"https://www.kaggle.com/pestipeti/fast-ensemble-5-folds-20-minutes\">this kernel</a></p>",
      "rawMarkdown": "It depends. I think 5-10 models/folds at least, if you optimize your inference kernel, probably you can ensemble more.\nTake a look at [this kernel](https://www.kaggle.com/pestipeti/fast-ensemble-5-folds-20-minutes)",
      "votes": null
    },
    {
      "id": "763216",
      "postDate": "03/04/2020 08:57:45",
      "content": "<p>Thank you for sharing your thoughts and the informative kernel :)</p>",
      "rawMarkdown": "Thank you for sharing your thoughts and the informative kernel :)",
      "votes": null
    },
    {
      "id": "763864",
      "postDate": "03/04/2020 23:04:04",
      "content": "<p>Subject to this constraint: <a href=\"https://www.kaggle.com/c/bengaliai-cv19/overview/notebooks-requirements\">https://www.kaggle.com/c/bengaliai-cv19/overview/notebooks-requirements</a>\nas many as you want :P</p>",
      "rawMarkdown": "Subject to this constraint: https://www.kaggle.com/c/bengaliai-cv19/overview/notebooks-requirements\nas many as you want :P",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 762452,
      "author_name": "pestipeti",
      "author_url": "",
      "post_date": "03/03/2020 14:20:50",
      "content": "<p>It depends. I think 5-10 models/folds at least, if you optimize your inference kernel, probably you can ensemble more.\nTake a look at <a href=\"https://www.kaggle.com/pestipeti/fast-ensemble-5-folds-20-minutes\">this kernel</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 763216,
          "author_name": "tanakin",
          "author_url": "",
          "post_date": "03/04/2020 08:57:45",
          "content": "<p>Thank you for sharing your thoughts and the informative kernel :)</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 763864,
      "author_name": "yuanlin08",
      "author_url": "",
      "post_date": "03/04/2020 23:04:04",
      "content": "<p>Subject to this constraint: <a href=\"https://www.kaggle.com/c/bengaliai-cv19/overview/notebooks-requirements\">https://www.kaggle.com/c/bengaliai-cv19/overview/notebooks-requirements</a>\nas many as you want :P</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "762410": "I try to run inferences of multiple models to ensemble the outputs in a submission kernel.\nBut it is hard to run long, or heavy inferences cause it may occur submission errors (i.e. Notebook Exceeded Allowed Compute).\nDid anyone check the maximum number of models possible in one-time kernel?",
    "762452": "It depends. I think 5-10 models/folds at least, if you optimize your inference kernel, probably you can ensemble more.\nTake a look at [this kernel](https://www.kaggle.com/pestipeti/fast-ensemble-5-folds-20-minutes)",
    "763216": "Thank you for sharing your thoughts and the informative kernel :)",
    "763864": "Subject to this constraint: https://www.kaggle.com/c/bengaliai-cv19/overview/notebooks-requirements\nas many as you want :P"
  },
  "source": "meta"
}