I have noticed one thing here, even though my submission takes 10-20 minutes on top, the scoring time is usually around 1 hour or so.
\nIs this always like that in code competitions? Or there is a way to speed up the API process?
", "messages": [ { "id": "2138017", "postDate": "02/10/2023 13:55:28", "content": "I have noticed one thing here, even though my submission takes 10-20 minutes on top, the scoring time is usually around 1 hour or so.
\nIs this always like that in code competitions? Or there is a way to speed up the API process?
", "rawMarkdown": "I have noticed one thing here, even though my submission takes 10-20 minutes on top, the scoring time is usually around 1 hour or so.\n\nIs this always like that in code competitions? Or there is a way to speed up the API process?", "votes": null }, { "id": "2138029", "postDate": "02/10/2023 14:00:20", "content": "I hope they will speed up the API process..
", "rawMarkdown": "I hope they will speed up the API process..", "votes": null }, { "id": "2138034", "postDate": "02/10/2023 14:04:00", "content": "yeah, it usually takes too much for me too….
", "rawMarkdown": "yeah, it usually takes too much for me too....", "votes": null }, { "id": "2138249", "postDate": "02/10/2023 17:02:24", "content": "Kaggle API is already fast.
", "rawMarkdown": "Kaggle API is already fast.", "votes": null }, { "id": "2186832", "postDate": "03/18/2023 06:23:14", "content": "my submission time takes 3-5minutes but scoring time is around 2.5 hour, i dont know why
", "rawMarkdown": "my submission time takes 3-5minutes but scoring time is around 2.5 hour, i dont know why", "votes": null }, { "id": "2187376", "postDate": "03/18/2023 17:21:36", "content": "Hi @shiyusha , I think the submission is on the mock test data (the small test we see when we run in interactive session). Then kaggle reruns the inference code on a different environment with a large test data and this is when the scoring occurs. I guess the reason it takes so long is that the feature engineering is called at each loop. During the training, we have all data at once and Feature Engineering is called only once.
", "rawMarkdown": "Hi @shiyusha , I think the submission is on the mock test data (the small test we see when we run in interactive session). Then kaggle reruns the inference code on a different environment with a large test data and this is when the scoring occurs. I guess the reason it takes so long is that the feature engineering is called at each loop. During the training, we have all data at once and Feature Engineering is called only once.", "votes": null }, { "id": "2187710", "postDate": "03/19/2023 01:21:29", "content": "Thank you very much for your reply, I think so, the hidden test set is large, and feature engineering functions are run in each iteration. inference time is very long.
", "rawMarkdown": "Thank you very much for your reply, I think so, the hidden test set is large, and feature engineering functions are run in each iteration. inference time is very long.", "votes": null } ], "comments": [ { "id": 2138029, "author_name": "karthiavenger45", "author_url": "", "post_date": "02/10/2023 14:00:20", "content": "I hope they will speed up the API process..
", "votes": null, "replies": [] }, { "id": 2138034, "author_name": "yogeshkumarsoni", "author_url": "", "post_date": "02/10/2023 14:04:00", "content": "yeah, it usually takes too much for me too….
", "votes": null, "replies": [] }, { "id": 2138249, "author_name": "titericz", "author_url": "", "post_date": "02/10/2023 17:02:24", "content": "Kaggle API is already fast.
", "votes": null, "replies": [] }, { "id": 2186832, "author_name": "shiyusha", "author_url": "", "post_date": "03/18/2023 06:23:14", "content": "my submission time takes 3-5minutes but scoring time is around 2.5 hour, i dont know why
", "votes": null, "replies": [ { "id": 2187376, "author_name": "gehallak", "author_url": "", "post_date": "03/18/2023 17:21:36", "content": "Hi @shiyusha , I think the submission is on the mock test data (the small test we see when we run in interactive session). Then kaggle reruns the inference code on a different environment with a large test data and this is when the scoring occurs. I guess the reason it takes so long is that the feature engineering is called at each loop. During the training, we have all data at once and Feature Engineering is called only once.
", "votes": null, "replies": [ { "id": 2187710, "author_name": "shiyusha", "author_url": "", "post_date": "03/19/2023 01:21:29", "content": "Thank you very much for your reply, I think so, the hidden test set is large, and feature engineering functions are run in each iteration. inference time is very long.
", "votes": null, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "2138017": "I have noticed one thing here, even though my submission takes 10-20 minutes on top, the scoring time is usually around 1 hour or so.\n\nIs this always like that in code competitions? Or there is a way to speed up the API process?", "2138029": "I hope they will speed up the API process..", "2138034": "yeah, it usually takes too much for me too....", "2138249": "Kaggle API is already fast.", "2186832": "my submission time takes 3-5minutes but scoring time is around 2.5 hour, i dont know why", "2187376": "Hi @shiyusha , I think the submission is on the mock test data (the small test we see when we run in interactive session). Then kaggle reruns the inference code on a different environment with a large test data and this is when the scoring occurs. I guess the reason it takes so long is that the feature engineering is called at each loop. During the training, we have all data at once and Feature Engineering is called only once.", "2187710": "Thank you very much for your reply, I think so, the hidden test set is large, and feature engineering functions are run in each iteration. inference time is very long." }, "source": "meta" }