i submitted a notebook include train model (not load the saved model), connect the kaggle server and run the test data predictions and return to the server. Normally, I run over this notebook locally took few minutes. But scoring takes normally few hours. So I am not sure which part cause this huge difference, i guess I am supposed to load a saved model,rather than train model because each batch data feed in, I was trainning a repeated model by using the local train data?
", "messages": [ { "id": "3064794", "postDate": "12/06/2024 02:25:10", "content": "i submitted a notebook include train model (not load the saved model), connect the kaggle server and run the test data predictions and return to the server. Normally, I run over this notebook locally took few minutes. But scoring takes normally few hours. So I am not sure which part cause this huge difference, i guess I am supposed to load a saved model,rather than train model because each batch data feed in, I was trainning a repeated model by using the local train data?
", "rawMarkdown": "i submitted a notebook include train model (not load the saved model), connect the kaggle server and run the test data predictions and return to the server. Normally, I run over this notebook locally took few minutes. But scoring takes normally few hours. So I am not sure which part cause this huge difference, i guess I am supposed to load a saved model,rather than train model because each batch data feed in, I was trainning a repeated model by using the local train data?", "votes": null }, { "id": "3065318", "postDate": "12/06/2024 17:16:38", "content": "Same problem here. I’m trying to understand how to use the evaluation API and just submitted my notebook, but I approached it the other way around. I trained the model on my PC and then uploaded it to Kaggle for submission, but it still takes ages for scoring. I’m wondering if the issue lies with my code or the server.
", "rawMarkdown": "Same problem here. I’m trying to understand how to use the evaluation API and just submitted my notebook, but I approached it the other way around. I trained the model on my PC and then uploaded it to Kaggle for submission, but it still takes ages for scoring. I’m wondering if the issue lies with my code or the server.", "votes": null }, { "id": "3065359", "postDate": "12/06/2024 17:50:07", "content": "Yes, save the model, then upload it as input to the notebook, make it so that your training cell of your notebook has if run_type == \"interactive\" so that it will only run if you interact with it not when you submit it.
", "rawMarkdown": "Yes, save the model, then upload it as input to the notebook, make it so that your training cell of your notebook has if run_type == \"interactive\" so that it will only run if you interact with it not when you submit it.", "votes": null }, { "id": "3067278", "postDate": "12/09/2024 06:03:05", "content": "thanks your reply
", "rawMarkdown": "thanks your reply", "votes": null }, { "id": "3067282", "postDate": "12/09/2024 06:05:17", "content": "sounds weird. In this way, rerun our codes do not need to train the model again. it should take less time to score.
", "rawMarkdown": "sounds weird. In this way, rerun our codes do not need to train the model again. it should take less time to score.", "votes": null } ], "comments": [ { "id": 3065318, "author_name": "ccyhui", "author_url": "", "post_date": "12/06/2024 17:16:38", "content": "Same problem here. I’m trying to understand how to use the evaluation API and just submitted my notebook, but I approached it the other way around. I trained the model on my PC and then uploaded it to Kaggle for submission, but it still takes ages for scoring. I’m wondering if the issue lies with my code or the server.
", "votes": null, "replies": [ { "id": 3067282, "author_name": "unknowntrader", "author_url": "", "post_date": "12/09/2024 06:05:17", "content": "sounds weird. In this way, rerun our codes do not need to train the model again. it should take less time to score.
", "votes": null, "replies": [] } ] }, { "id": 3065359, "author_name": "guillaumecollin12", "author_url": "", "post_date": "12/06/2024 17:50:07", "content": "Yes, save the model, then upload it as input to the notebook, make it so that your training cell of your notebook has if run_type == \"interactive\" so that it will only run if you interact with it not when you submit it.
", "votes": null, "replies": [ { "id": 3067278, "author_name": "unknowntrader", "author_url": "", "post_date": "12/09/2024 06:03:05", "content": "thanks your reply
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "3064794": "i submitted a notebook include train model (not load the saved model), connect the kaggle server and run the test data predictions and return to the server. Normally, I run over this notebook locally took few minutes. But scoring takes normally few hours. So I am not sure which part cause this huge difference, i guess I am supposed to load a saved model,rather than train model because each batch data feed in, I was trainning a repeated model by using the local train data?", "3065318": "Same problem here. I’m trying to understand how to use the evaluation API and just submitted my notebook, but I approached it the other way around. I trained the model on my PC and then uploaded it to Kaggle for submission, but it still takes ages for scoring. I’m wondering if the issue lies with my code or the server.", "3065359": "Yes, save the model, then upload it as input to the notebook, make it so that your training cell of your notebook has if run_type == \"interactive\" so that it will only run if you interact with it not when you submit it.", "3067278": "thanks your reply", "3067282": "sounds weird. In this way, rerun our codes do not need to train the model again. it should take less time to score." }, "source": "meta" }