{
  "id": 614186,
  "title": "Are offline pretrained models allowed in this competition?",
  "url": "/competitions/csiro-biomass/discussion/614186",
  "author_name": "",
  "post_date": "2025-11-01T18:57:04.376277Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi everyone,</p>\n<p>I had a quick question regarding the model usage policy for this competition. Since the notebook environment doesn’t have internet access during runtime, are we allowed to use pretrained neural network models (for example, downloaded externally and added to the input folder), or are we expected to train our own models from scratch within the provided dataset?</p>\n<p>I just want to confirm whether using offline pretrained weights or models (e.g., ResNet, EfficientNet, etc.) is allowed as long as they are uploaded with the notebook, or if the competition rules require building and training models entirely from scratch ?</p>",
  "messages": [
    {
      "id": "3309947",
      "postDate": "11/01/2025 18:57:04",
      "content": "<p>Hi everyone,</p>\n<p>I had a quick question regarding the model usage policy for this competition. Since the notebook environment doesn’t have internet access during runtime, are we allowed to use pretrained neural network models (for example, downloaded externally and added to the input folder), or are we expected to train our own models from scratch within the provided dataset?</p>\n<p>I just want to confirm whether using offline pretrained weights or models (e.g., ResNet, EfficientNet, etc.) is allowed as long as they are uploaded with the notebook, or if the competition rules require building and training models entirely from scratch ?</p>",
      "rawMarkdown": "Hi everyone,\n\nI had a quick question regarding the model usage policy for this competition. Since the notebook environment doesn’t have internet access during runtime, are we allowed to use pretrained neural network models (for example, downloaded externally and added to the input folder), or are we expected to train our own models from scratch within the provided dataset?\n\nI just want to confirm whether using offline pretrained weights or models (e.g., ResNet, EfficientNet, etc.) is allowed as long as they are uploaded with the notebook, or if the competition rules require building and training models entirely from scratch ?",
      "votes": null
    },
    {
      "id": "3310034",
      "postDate": "11/01/2025 23:13:27",
      "content": "<p>You can use any model - there are lots of examples in other competitions on how to use a dataset for loading models.  I train all my models on local machines and than put the saved model in a dataset, than add the dataset to the prediction notebook.</p>",
      "rawMarkdown": "You can use any model - there are lots of examples in other competitions on how to use a dataset for loading models.  I train all my models on local machines and than put the saved model in a dataset, than add the dataset to the prediction notebook.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3310034,
      "author_name": "pcjimmmy",
      "author_url": "",
      "post_date": "11/01/2025 23:13:27",
      "content": "<p>You can use any model - there are lots of examples in other competitions on how to use a dataset for loading models.  I train all my models on local machines and than put the saved model in a dataset, than add the dataset to the prediction notebook.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3309947": "Hi everyone,\n\nI had a quick question regarding the model usage policy for this competition. Since the notebook environment doesn’t have internet access during runtime, are we allowed to use pretrained neural network models (for example, downloaded externally and added to the input folder), or are we expected to train our own models from scratch within the provided dataset?\n\nI just want to confirm whether using offline pretrained weights or models (e.g., ResNet, EfficientNet, etc.) is allowed as long as they are uploaded with the notebook, or if the competition rules require building and training models entirely from scratch ?",
    "3310034": "You can use any model - there are lots of examples in other competitions on how to use a dataset for loading models.  I train all my models on local machines and than put the saved model in a dataset, than add the dataset to the prediction notebook."
  },
  "source": "meta"
}