Hi Everyone,\nThis is my first time participating in Kaggle competition and also I'm new to machine learning as well.\nI did not understand how to create an inference kernel which can be used to just run on test data and in stage 2 or private test data.\nMy code for training and generating submission.csv is ready but it takes several hours to train. If someone please guide me with steps (and small code section) to dump my trained model as dataset and then upload it in a new kernel and use it for submission, it would be very helpful.\nThanks.
", "messages": [ { "id": "547037", "postDate": "06/07/2019 07:20:28", "content": "Hi Everyone,\nThis is my first time participating in Kaggle competition and also I'm new to machine learning as well.\nI did not understand how to create an inference kernel which can be used to just run on test data and in stage 2 or private test data.\nMy code for training and generating submission.csv is ready but it takes several hours to train. If someone please guide me with steps (and small code section) to dump my trained model as dataset and then upload it in a new kernel and use it for submission, it would be very helpful.\nThanks.
", "rawMarkdown": "Hi Everyone,\nThis is my first time participating in Kaggle competition and also I'm new to machine learning as well.\nI did not understand how to create an inference kernel which can be used to just run on test data and in stage 2 or private test data.\nMy code for training and generating submission.csv is ready but it takes several hours to train. If someone please guide me with steps (and small code section) to dump my trained model as dataset and then upload it in a new kernel and use it for submission, it would be very helpful.\nThanks.", "votes": null }, { "id": "547060", "postDate": "06/07/2019 08:04:10", "content": "You can save your best model or set of models in your training kernel and commit. Once commit is done download the model from the output section of your kernel. Create a dataset and upload your model. In your inference kernel add this dataset, know your model is available and can be used for prediction.
\n\nPlease note that u would need to preprocess your test set data in the inference kernel as test data set will be new and 3 times the size as current dataset.\nCheck out and upvote this kernel if useful\nhttps://www.kaggle.com/chandramani/sample-inference-submissionkernel
", "rawMarkdown": "You can save your best model or set of models in your training kernel and commit. Once commit is done download the model from the output section of your kernel. Create a dataset and upload your model. In your inference kernel add this dataset, know your model is available and can be used for prediction.\n\nPlease note that u would need to preprocess your test set data in the inference kernel as test data set will be new and 3 times the size as current dataset.\nCheck out and upvote this kernel if useful\nhttps://www.kaggle.com/chandramani/sample-inference-submissionkernel", "votes": null }, { "id": "547194", "postDate": "06/07/2019 11:43:10", "content": "Hi Mani,\nThank you very much, this is very helpful.\nI'm going through it and will let know if I've any questions.\nThanks again for your awesome kernel!!
", "rawMarkdown": "Hi Mani,\nThank you very much, this is very helpful.\nI'm going through it and will let know if I've any questions.\nThanks again for your awesome kernel!!", "votes": null }, { "id": "547201", "postDate": "06/07/2019 11:59:16", "content": "Happy to help
", "rawMarkdown": "Happy to help", "votes": null } ], "comments": [ { "id": 547060, "author_name": "chandramani", "author_url": "", "post_date": "06/07/2019 08:04:10", "content": "You can save your best model or set of models in your training kernel and commit. Once commit is done download the model from the output section of your kernel. Create a dataset and upload your model. In your inference kernel add this dataset, know your model is available and can be used for prediction.
\n\nPlease note that u would need to preprocess your test set data in the inference kernel as test data set will be new and 3 times the size as current dataset.\nCheck out and upvote this kernel if useful\nhttps://www.kaggle.com/chandramani/sample-inference-submissionkernel
", "votes": null, "replies": [ { "id": 547194, "author_name": "ackusingh", "author_url": "", "post_date": "06/07/2019 11:43:10", "content": "Hi Mani,\nThank you very much, this is very helpful.\nI'm going through it and will let know if I've any questions.\nThanks again for your awesome kernel!!
", "votes": null, "replies": [] }, { "id": 547201, "author_name": "chandramani", "author_url": "", "post_date": "06/07/2019 11:59:16", "content": "Happy to help
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "547037": "Hi Everyone,\nThis is my first time participating in Kaggle competition and also I'm new to machine learning as well.\nI did not understand how to create an inference kernel which can be used to just run on test data and in stage 2 or private test data.\nMy code for training and generating submission.csv is ready but it takes several hours to train. If someone please guide me with steps (and small code section) to dump my trained model as dataset and then upload it in a new kernel and use it for submission, it would be very helpful.\nThanks.", "547060": "You can save your best model or set of models in your training kernel and commit. Once commit is done download the model from the output section of your kernel. Create a dataset and upload your model. In your inference kernel add this dataset, know your model is available and can be used for prediction.\n\nPlease note that u would need to preprocess your test set data in the inference kernel as test data set will be new and 3 times the size as current dataset.\nCheck out and upvote this kernel if useful\nhttps://www.kaggle.com/chandramani/sample-inference-submissionkernel", "547194": "Hi Mani,\nThank you very much, this is very helpful.\nI'm going through it and will let know if I've any questions.\nThanks again for your awesome kernel!!", "547201": "Happy to help" }, "source": "meta" }