{
  "id": 411812,
  "title": "timeout error and CPU speed stability",
  "url": "/competitions/birdclef-2023/discussion/411812",
  "author_name": "",
  "post_date": "2023-05-21T07:11:19.739417800Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Today, our team suddenly started to have timeout errors although we are using the same inference code and the same model architecture as before (we haven't got any errors so far). The only difference is model weights. <br>\nIs Kaggle CPU power stable ? Does inference speed vary a lot?</p>",
  "messages": [
    {
      "id": "2267710",
      "postDate": "05/21/2023 07:11:19",
      "content": "<p>Today, our team suddenly started to have timeout errors although we are using the same inference code and the same model architecture as before (we haven't got any errors so far). The only difference is model weights. <br>\nIs Kaggle CPU power stable ? Does inference speed vary a lot?</p>",
      "rawMarkdown": "Today, our team suddenly started to have timeout errors although we are using the same inference code and the same model architecture as before (we haven't got any errors so far). The only difference is model weights. \nIs Kaggle CPU power stable ? Does inference speed vary a lot?",
      "votes": null
    },
    {
      "id": "2268096",
      "postDate": "05/21/2023 12:55:26",
      "content": "<p>CPU matters, but model weights also matters….So maybe it is good to make inference be around 115min<br>\nI am also striving to avoid timeout.</p>",
      "rawMarkdown": "CPU matters, but model weights also matters....So maybe it is good to make inference be around 115min\nI am also striving to avoid timeout.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2268096,
      "author_name": "honglihang",
      "author_url": "",
      "post_date": "05/21/2023 12:55:26",
      "content": "<p>CPU matters, but model weights also matters….So maybe it is good to make inference be around 115min<br>\nI am also striving to avoid timeout.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2267710": "Today, our team suddenly started to have timeout errors although we are using the same inference code and the same model architecture as before (we haven't got any errors so far). The only difference is model weights. \nIs Kaggle CPU power stable ? Does inference speed vary a lot?",
    "2268096": "CPU matters, but model weights also matters....So maybe it is good to make inference be around 115min\nI am also striving to avoid timeout."
  },
  "source": "meta"
}