The model ran very well on the Internet, but I faced an error after sending it.
", "messages": [ { "id": "3152878", "postDate": "03/18/2025 08:24:09", "content": "The model ran very well on the Internet, but I faced an error after sending it.
", "rawMarkdown": "The model ran very well on the Internet, but I faced an error after sending it.", "votes": null }, { "id": "3153402", "postDate": "03/18/2025 19:05:44", "content": "What type of error? The internet has to be turned off on your notebook when you submit so I am guessing that when you are loading a model you are trying to download the pretrained weights and it fails because it runs offline. If you are using something like ResNet then make sure you pass pretrained=False so it does not try and download weights.
", "rawMarkdown": "What type of error? The internet has to be turned off on your notebook when you submit so I am guessing that when you are loading a model you are trying to download the pretrained weights and it fails because it runs offline. If you are using something like ResNet then make sure you pass pretrained=False so it does not try and download weights.", "votes": null }, { "id": "3153701", "postDate": "03/19/2025 04:41:12", "content": "I turned off the internet and submitted the notebook. I am using something like the ResNet model. I trained the model on the same notebook, and inference it.
", "rawMarkdown": "I turned off the internet and submitted the notebook. I am using something like the ResNet model. I trained the model on the same notebook, and inference it.", "votes": null }, { "id": "3154172", "postDate": "03/19/2025 15:48:47", "content": "I recommend training on one notebook, saving the weights, and then loading them into your inference notebook by using a kaggle dataset to store the weights. This reduces the submission times and removes some randomness as your notebook will be rerun on the final dataset and you do not want it to perform worse a second time.
", "rawMarkdown": "I recommend training on one notebook, saving the weights, and then loading them into your inference notebook by using a kaggle dataset to store the weights. This reduces the submission times and removes some randomness as your notebook will be rerun on the final dataset and you do not want it to perform worse a second time.", "votes": null }, { "id": "3154458", "postDate": "03/20/2025 00:56:14", "content": "Thank you. But I don't know how to do it? How can I load model from another notebook? Could you give me instruction?
", "rawMarkdown": "Thank you. But I don't know how to do it? How can I load model from another notebook? Could you give me instruction?", "votes": null }, { "id": "3154591", "postDate": "03/20/2025 05:59:40", "content": "@connorjd If you store a model as a Kaggle dataset, you should make it public, right?
", "rawMarkdown": "connorjd If you store a model as a Kaggle dataset, you should make it public, right?", "votes": null }, { "id": "3156109", "postDate": "03/21/2025 18:28:57", "content": "In the starter code, I demonstrate how to do this. The weights in train yolo are loaded into the submission notebook. You should be able to copy this approach in pytorch or tensorflow or really anything that you're using.
\nhttps://www.kaggle.com/code/andrewjdarley/byu-biophysics-group-starter
", "rawMarkdown": "In the starter code, I demonstrate how to do this. The weights in train yolo are loaded into the submission notebook. You should be able to copy this approach in pytorch or tensorflow or really anything that you're using. \n\nhttps://www.kaggle.com/code/andrewjdarley/byu-biophysics-group-starter", "votes": null } ], "comments": [ { "id": 3153402, "author_name": "connorjd", "author_url": "", "post_date": "03/18/2025 19:05:44", "content": "What type of error? The internet has to be turned off on your notebook when you submit so I am guessing that when you are loading a model you are trying to download the pretrained weights and it fails because it runs offline. If you are using something like ResNet then make sure you pass pretrained=False so it does not try and download weights.
", "votes": null, "replies": [ { "id": 3153701, "author_name": "rustambazarbayev", "author_url": "", "post_date": "03/19/2025 04:41:12", "content": "I turned off the internet and submitted the notebook. I am using something like the ResNet model. I trained the model on the same notebook, and inference it.
", "votes": null, "replies": [ { "id": 3154172, "author_name": "connorjd", "author_url": "", "post_date": "03/19/2025 15:48:47", "content": "I recommend training on one notebook, saving the weights, and then loading them into your inference notebook by using a kaggle dataset to store the weights. This reduces the submission times and removes some randomness as your notebook will be rerun on the final dataset and you do not want it to perform worse a second time.
", "votes": null, "replies": [ { "id": 3154458, "author_name": "rustambazarbayev", "author_url": "", "post_date": "03/20/2025 00:56:14", "content": "Thank you. But I don't know how to do it? How can I load model from another notebook? Could you give me instruction?
", "votes": null, "replies": [ { "id": 3156109, "author_name": "andrewjdarley", "author_url": "", "post_date": "03/21/2025 18:28:57", "content": "In the starter code, I demonstrate how to do this. The weights in train yolo are loaded into the submission notebook. You should be able to copy this approach in pytorch or tensorflow or really anything that you're using.
\nhttps://www.kaggle.com/code/andrewjdarley/byu-biophysics-group-starter
", "votes": null, "replies": [] } ] }, { "id": 3154591, "author_name": "alekslinya", "author_url": "", "post_date": "03/20/2025 05:59:40", "content": "@connorjd If you store a model as a Kaggle dataset, you should make it public, right?
", "votes": null, "replies": [] } ] } ] } ] } ], "raw_markdown_by_id": { "3152878": "The model ran very well on the Internet, but I faced an error after sending it.", "3153402": "What type of error? The internet has to be turned off on your notebook when you submit so I am guessing that when you are loading a model you are trying to download the pretrained weights and it fails because it runs offline. If you are using something like ResNet then make sure you pass pretrained=False so it does not try and download weights.", "3153701": "I turned off the internet and submitted the notebook. I am using something like the ResNet model. I trained the model on the same notebook, and inference it.", "3154172": "I recommend training on one notebook, saving the weights, and then loading them into your inference notebook by using a kaggle dataset to store the weights. This reduces the submission times and removes some randomness as your notebook will be rerun on the final dataset and you do not want it to perform worse a second time.", "3154458": "Thank you. But I don't know how to do it? How can I load model from another notebook? Could you give me instruction?", "3154591": "connorjd If you store a model as a Kaggle dataset, you should make it public, right?", "3156109": "In the starter code, I demonstrate how to do this. The weights in train yolo are loaded into the submission notebook. You should be able to copy this approach in pytorch or tensorflow or really anything that you're using. \n\nhttps://www.kaggle.com/code/andrewjdarley/byu-biophysics-group-starter" }, "source": "meta" }