I seems not to get, what the competitors mean, by this statement. Any one to show me a practical example, will be highly appreciated. Thanks.
", "messages": [ { "id": 2486992, "postDate": "2023-10-18T09:37:36.087Z", "content": "This means, that you have to create two separate notebooks. One for training the model, and one for inference from the trained model.
\nA basic workflow would be - Make a notebook, develop a model and train it in the notebook, and save it as a binary file.
\nMake a second notebook, and import the training notebook as a data source and load the model. Use the loaded model to infer on the test data, and submit to the competition. Here are a few examples from this competition:
By pjmathematician (me):
\nTraining : https://www.kaggle.com/code/pjmathematician/ubco-keras-cnn-baseline-thumbnails-training
\nInference : https://www.kaggle.com/code/pjmathematician/ubco-keras-cnn-baseline-thumbnails-inference
By motono0223 :
\nTraining : https://www.kaggle.com/code/motono0223/ubc-pytorch-cnn-training-fold1of5
\nInference : https://www.kaggle.com/code/motono0223/ubc-pytorch-cnn-inference
By JIRKA BOROVEC :
\nTraining : https://www.kaggle.com/code/jirkaborovec/cancer-subtype-baseline-with-lightning-timm
\nInference : https://www.kaggle.com/code/jirkaborovec/cancer-subtype-lightning-torch-inference
Hope this helps! Goodluck for the competition!
", "rawMarkdown": "This means, that you have to create two separate notebooks. One for training the model, and one for inference from the trained model.\nA basic workflow would be - Make a notebook, develop a model and train it in the notebook, and save it as a binary file. \nMake a second notebook, and import the training notebook as a data source and load the model. Use the loaded model to infer on the test data, and submit to the competition. Here are a few examples from this competition:\n\n1. By pjmathematician (me):\nTraining : https://www.kaggle.com/code/pjmathematician/ubco-keras-cnn-baseline-thumbnails-training\nInference : https://www.kaggle.com/code/pjmathematician/ubco-keras-cnn-baseline-thumbnails-inference\n\n2. By motono0223 : \nTraining : https://www.kaggle.com/code/motono0223/ubc-pytorch-cnn-training-fold1of5\nInference : https://www.kaggle.com/code/motono0223/ubc-pytorch-cnn-inference\n\n3. By JIRKA BOROVEC : \nTraining : https://www.kaggle.com/code/jirkaborovec/cancer-subtype-baseline-with-lightning-timm\nInference : https://www.kaggle.com/code/jirkaborovec/cancer-subtype-lightning-torch-inference\n\nHope this helps! Goodluck for the competition!", "votes": 5, "replies": [ { "id": 2487129, "postDate": "2023-10-18T11:43:36.397Z", "content": "Thanks so much, for the guide.
", "rawMarkdown": "Thanks so much, for the guide.", "replies": [ { "id": 2488530, "postDate": "2023-10-19T09:57:39.277Z", "content": "I trained my model on my own server and load it as a dataset.There is still a large gape between RTX4090
\nand P100.
\nI think it can save a lot of time.
Yeah， My server is just fine，but I got a “threw exception” error in kaggle😭
", "rawMarkdown": "Yeah， My server is just fine，but I got a “threw exception” error in kaggle😭" }, { "id": 2488616, "postDate": "2023-10-19T11:25:32.603Z", "rawMarkdown": "", "isDeleted": true }, { "id": 2488741, "postDate": "2023-10-19T13:10:52.953Z", "content": "Yes,after waiting for 2hours I get “Notebook Out of Memory”,you know it is really terrible.
\nI'm trying to use this idea in my submissionhttps://www.kaggle.com/code/sunilthite/ubc-ocean-testing-data-prediction,but I faced “Notebook Out of Memory”.
I seems not to get, what the competitors mean, by this statement. Any one to show me a practical example, will be highly appreciated. Thanks.
", "rawMarkdown": "I seems not to get, what the competitors mean, by this statement. Any one to show me a practical example, will be highly appreciated. Thanks.", "votes": 1 } ], "comments": [ { "id": 2486992, "author_name": "pjmathematician", "author_url": "", "post_date": "2023-10-18T09:37:36.087000", "content": "This means, that you have to create two separate notebooks. One for training the model, and one for inference from the trained model.
\nA basic workflow would be - Make a notebook, develop a model and train it in the notebook, and save it as a binary file.
\nMake a second notebook, and import the training notebook as a data source and load the model. Use the loaded model to infer on the test data, and submit to the competition. Here are a few examples from this competition:
By pjmathematician (me):
\nTraining : https://www.kaggle.com/code/pjmathematician/ubco-keras-cnn-baseline-thumbnails-training
\nInference : https://www.kaggle.com/code/pjmathematician/ubco-keras-cnn-baseline-thumbnails-inference
By motono0223 :
\nTraining : https://www.kaggle.com/code/motono0223/ubc-pytorch-cnn-training-fold1of5
\nInference : https://www.kaggle.com/code/motono0223/ubc-pytorch-cnn-inference
By JIRKA BOROVEC :
\nTraining : https://www.kaggle.com/code/jirkaborovec/cancer-subtype-baseline-with-lightning-timm
\nInference : https://www.kaggle.com/code/jirkaborovec/cancer-subtype-lightning-torch-inference
Hope this helps! Goodluck for the competition!
", "votes": 5, "replies": [ { "id": 2487129, "author_name": "Raheem Nasirudeen", "author_url": "", "post_date": "2023-10-18T11:43:36.397000", "content": "Thanks so much, for the guide.
", "votes": 0, "replies": [ { "id": 2488530, "author_name": "LuoZiqian", "author_url": "", "post_date": "2023-10-19T09:57:39.277000", "content": "I trained my model on my own server and load it as a dataset.There is still a large gape between RTX4090
\nand P100.
\nI think it can save a lot of time.
Yeah， My server is just fine，but I got a “threw exception” error in kaggle😭
", "votes": 0, "replies": [] }, { "id": 2488616, "author_name": "", "author_url": "", "post_date": "2023-10-19T11:25:32.603000", "content": "", "votes": 0, "replies": [] }, { "id": 2488741, "author_name": "LuoZiqian", "author_url": "", "post_date": "2023-10-19T13:10:52.953000", "content": "Yes,after waiting for 2hours I get “Notebook Out of Memory”,you know it is really terrible.
\nI'm trying to use this idea in my submissionhttps://www.kaggle.com/code/sunilthite/ubc-ocean-testing-data-prediction,but I faced “Notebook Out of Memory”.