{
  "id": 301133,
  "title": "How to speed up keras model.predict ",
  "url": "/competitions/g-research-crypto-forecasting/discussion/301133",
  "author_name": "cymchen",
  "post_date": "2022-01-16T03:39:29.552000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I use MLP to predict futures prices. In my local I tried 5folds*14assets and it performs pretty well. But in notebook it would raise memory error for 70 models and runtime error, is there anyway to solve thoses problems ?   Or the onlything I can do is to reduce the model numbers?</p>",
  "messages": [
    {
      "id": 1651748,
      "postDate": "2022-01-16T03:39:29.553Z",
      "content": "<p>I use MLP to predict futures prices. In my local I tried 5folds*14assets and it performs pretty well. But in notebook it would raise memory error for 70 models and runtime error, is there anyway to solve thoses problems ?   Or the onlything I can do is to reduce the model numbers?</p>",
      "rawMarkdown": "I use MLP to predict futures prices. In my local I tried 5folds*14assets and it performs pretty well. But in notebook it would raise memory error for 70 models and runtime error, is there anyway to solve thoses problems ?   Or the onlything I can do is to reduce the model numbers?"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1651748": "I use MLP to predict futures prices. In my local I tried 5folds*14assets and it performs pretty well. But in notebook it would raise memory error for 70 models and runtime error, is there anyway to solve thoses problems ?   Or the onlything I can do is to reduce the model numbers?"
  }
}