Hello,
\nI am using a standard jupyter notebook to process the training phase locally, however, I'd like to be able to use the Kaggle evaluation API within the same notebook. Is this at all possible? 1.) I notice that a lot of the underlying code references to the /kaggle/input directory, and I'm not sure if it's as simple as changing it to my local environment. 2.) You need to submit a Kaggle notebook to evaluate the results, which presumably would be on Kaggle. Has anyone gotten this working?
", "messages": [ { "id": 3031535, "postDate": "2024-10-29T19:47:18.743Z", "content": "Yes, I have the same problem. This is a good notebook to start https://www.kaggle.com/code/shiyili/js24-rmf-submission-api-debug-with-synthetic-test. Thanks to @shiyili.
\nYou can prepare test set in the way it does, but then instead of using kaggle API one has to write some local implementation. The only thing is ensure that it behaves in the same way as the kaggle online one.
The actual evaluation needs to run on Kaggle but you can write a local implementation of it to speed up local testing and iteration.
\nE.g create a test holdout set from the training data and pass it in one time step at a time and then calculate the final R2 at the end of the test period
", "rawMarkdown": "The actual evaluation needs to run on Kaggle but you can write a local implementation of it to speed up local testing and iteration.\n\nE.g create a test holdout set from the training data and pass it in one time step at a time and then calculate the final R2 at the end of the test period", "votes": 1 }, { "id": 3029018, "postDate": "2024-10-26T18:50:33.953Z", "content": "Hello,
\nI am using a standard jupyter notebook to process the training phase locally, however, I'd like to be able to use the Kaggle evaluation API within the same notebook. Is this at all possible? 1.) I notice that a lot of the underlying code references to the /kaggle/input directory, and I'm not sure if it's as simple as changing it to my local environment. 2.) You need to submit a Kaggle notebook to evaluate the results, which presumably would be on Kaggle. Has anyone gotten this working?
", "rawMarkdown": "Hello, \n\nI am using a standard jupyter notebook to process the training phase locally, however, I'd like to be able to use the Kaggle evaluation API within the same notebook. Is this at all possible? 1.) I notice that a lot of the underlying code references to the /kaggle/input directory, and I'm not sure if it's as simple as changing it to my local environment. 2.) You need to submit a Kaggle notebook to evaluate the results, which presumably would be on Kaggle. Has anyone gotten this working? ", "votes": 1 }, { "id": 3029057, "postDate": "2024-10-26T19:50:35.877Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 3031535, "author_name": "Maxim Shatskiy", "author_url": "", "post_date": "2024-10-29T19:47:18.743000", "content": "Yes, I have the same problem. This is a good notebook to start https://www.kaggle.com/code/shiyili/js24-rmf-submission-api-debug-with-synthetic-test. Thanks to @shiyili.
\nYou can prepare test set in the way it does, but then instead of using kaggle API one has to write some local implementation. The only thing is ensure that it behaves in the same way as the kaggle online one.
The actual evaluation needs to run on Kaggle but you can write a local implementation of it to speed up local testing and iteration.
\nE.g create a test holdout set from the training data and pass it in one time step at a time and then calculate the final R2 at the end of the test period
", "votes": 1, "replies": [] }, { "id": 3029057, "author_name": "", "author_url": "", "post_date": "2024-10-26T19:50:35.877000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "3031535": "Yes, I have the same problem. This is a good notebook to start https://www.kaggle.com/code/shiyili/js24-rmf-submission-api-debug-with-synthetic-test. Thanks to @shiyili.\nYou can prepare test set in the way it does, but then instead of using kaggle API one has to write some local implementation. The only thing is ensure that it behaves in the same way as the kaggle online one.", "3029141": "The actual evaluation needs to run on Kaggle but you can write a local implementation of it to speed up local testing and iteration.\n\nE.g create a test holdout set from the training data and pass it in one time step at a time and then calculate the final R2 at the end of the test period", "3029018": "Hello, \n\nI am using a standard jupyter notebook to process the training phase locally, however, I'd like to be able to use the Kaggle evaluation API within the same notebook. Is this at all possible? 1.) I notice that a lot of the underlying code references to the /kaggle/input directory, and I'm not sure if it's as simple as changing it to my local environment. 2.) You need to submit a Kaggle notebook to evaluate the results, which presumably would be on Kaggle. Has anyone gotten this working? ", "3029057": "" } }