Could someone clarify the rules? They state:
\nThis is a Code Competition\nSubmissions to this competition must be made through Notebooks. In order for the \"Submit\" button to be active after a commit, the following conditions must be met:\n\nCPU Notebook <= 9 hours run-time\nGPU Notebook <= 9 hours run-time\nInternet access disabled\nFreely & publicly available external data is allowed, including pre-trained models\nIf we need to submit via a notebook, and we can't access the internet, how do we upload our external data?
Does 9 hours include training time? What if we train outside of the notebook, publish it freely, and then load it as \"free and public data\"?
I'm new to Kaggle, so I don't know if these questions are unique to this challenge or apply to others.
", "messages": [ { "id": 1855758, "postDate": "2022-07-14T21:49:01.757Z", "content": "Hi there,
\nYou can train offline and use the Kaggle notebook as just your inference code - however your training code may be requested in validating a model if you win.
", "rawMarkdown": "Hi there,\n\nYou can train offline and use the Kaggle notebook as just your inference code - however your training code may be requested in validating a model if you win.", "votes": 1, "replies": [ { "id": 1875469, "postDate": "2022-07-29T03:56:02.690Z", "content": "Can you clarify: It sounded like the submission needs to be in the form of a notebook only. How does that notebook get access to our offline-produced files? Do we upload those files to Kaggle somehow and then the notebook is allowed to read them?
\nWhat exactly does the rule \"submissions must be in the form of notebooks\" mean?
", "rawMarkdown": "Can you clarify: It sounded like the submission needs to be in the form of a notebook only. How does that notebook get access to our offline-produced files? Do we upload those files to Kaggle somehow and then the notebook is allowed to read them?\n\nWhat exactly does the rule \"submissions must be in the form of notebooks\" mean?" }, { "id": 1875719, "postDate": "2022-07-29T08:23:48.793Z", "content": "You can upload your model weights as a private dataset and add it to the notebook, you can check the public inference notebooks as an example.
", "rawMarkdown": "You can upload your model weights as a private dataset and add it to the notebook, you can check the public inference notebooks as an example." } ] }, { "id": 1855287, "postDate": "2022-07-14T12:33:18.867Z", "content": "You can upload the files you need as datasets, and with limitations (only the latest version is visible) notebook output - the recent cubert example (https://www.kaggle.com/code/nbroad/using-cubert/notebook) shows how to do this.
\nFine-tuned models that you produce (and can reproduce) using your own code do not have to be public datasets.
", "rawMarkdown": "You can upload the files you need as datasets, and with limitations (only the latest version is visible) notebook output - the recent cubert example (https://www.kaggle.com/code/nbroad/using-cubert/notebook) shows how to do this.\n\nFine-tuned models that you produce (and can reproduce) using your own code do *not* have to be public datasets." }, { "id": 1854712, "postDate": "2022-07-14T00:11:02.423Z", "content": "I'm new on Kaggle too. However, I think I'm apt to try answer those two question.
\nIn you kaggle notebook, you'll find an option in Settings that allow you to turn your internet on/off as shown in this.
\nThe runtime is for your entire notebook. Therefore, it should include training time too (though I'm not sure if it's possible to upload a trained model together the notebook).
", "rawMarkdown": "I'm new on Kaggle too. However, I think I'm apt to try answer those two question. \n\nIn you kaggle notebook, you'll find an option in Settings that allow you to turn your internet on/off as shown in [this](https://www.kaggle.com/product-feedback/63544?adlt=strict&toWww=1&redig=F584FCFDC7C2404AB5E49409C9908D4E). \n\nThe runtime is for your entire notebook. Therefore, it should include training time too (though I'm not sure if it's possible to upload a trained model together the notebook).\n\n\n" }, { "id": 1853232, "postDate": "2022-07-12T17:48:21.527Z", "content": "Could someone clarify the rules? They state:
\nThis is a Code Competition\nSubmissions to this competition must be made through Notebooks. In order for the \"Submit\" button to be active after a commit, the following conditions must be met:\n\nCPU Notebook <= 9 hours run-time\nGPU Notebook <= 9 hours run-time\nInternet access disabled\nFreely & publicly available external data is allowed, including pre-trained models\nIf we need to submit via a notebook, and we can't access the internet, how do we upload our external data?
Does 9 hours include training time? What if we train outside of the notebook, publish it freely, and then load it as \"free and public data\"?
I'm new to Kaggle, so I don't know if these questions are unique to this challenge or apply to others.
", "rawMarkdown": "Could someone clarify the rules? They state:\n```\nThis is a Code Competition\nSubmissions to this competition must be made through Notebooks. In order for the \"Submit\" button to be active after a commit, the following conditions must be met:\n\nCPU Notebook <= 9 hours run-time\nGPU Notebook <= 9 hours run-time\nInternet access disabled\nFreely & publicly available external data is allowed, including pre-trained models\n```\n\n1. If we need to submit via a notebook, and we can't access the internet, how do we upload our external data?\n\n2. Does 9 hours include training time? What if we train outside of the notebook, publish it freely, and then load it as \"free and public data\"?\n\nI'm new to Kaggle, so I don't know if these questions are unique to this challenge or apply to others." }, { "id": 1875350, "postDate": "2022-07-29T00:17:59.687Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 1855758, "author_name": "Addison Howard", "author_url": "", "post_date": "2022-07-14T21:49:01.757000", "content": "Hi there,
\nYou can train offline and use the Kaggle notebook as just your inference code - however your training code may be requested in validating a model if you win.
", "votes": 1, "replies": [ { "id": 1875469, "author_name": "S. Robert James", "author_url": "", "post_date": "2022-07-29T03:56:02.690000", "content": "Can you clarify: It sounded like the submission needs to be in the form of a notebook only. How does that notebook get access to our offline-produced files? Do we upload those files to Kaggle somehow and then the notebook is allowed to read them?
\nWhat exactly does the rule \"submissions must be in the form of notebooks\" mean?
", "votes": 0, "replies": [] }, { "id": 1875719, "author_name": "Dmytro Poplavskiy", "author_url": "", "post_date": "2022-07-29T08:23:48.793000", "content": "You can upload your model weights as a private dataset and add it to the notebook, you can check the public inference notebooks as an example.
", "votes": 0, "replies": [] } ] }, { "id": 1855287, "author_name": "happycube", "author_url": "", "post_date": "2022-07-14T12:33:18.867000", "content": "You can upload the files you need as datasets, and with limitations (only the latest version is visible) notebook output - the recent cubert example (https://www.kaggle.com/code/nbroad/using-cubert/notebook) shows how to do this.
\nFine-tuned models that you produce (and can reproduce) using your own code do not have to be public datasets.
", "votes": 0, "replies": [] }, { "id": 1854712, "author_name": "Disguiso", "author_url": "", "post_date": "2022-07-14T00:11:02.423000", "content": "I'm new on Kaggle too. However, I think I'm apt to try answer those two question.
\nIn you kaggle notebook, you'll find an option in Settings that allow you to turn your internet on/off as shown in this.
\nThe runtime is for your entire notebook. Therefore, it should include training time too (though I'm not sure if it's possible to upload a trained model together the notebook).
", "votes": 0, "replies": [] }, { "id": 1875350, "author_name": "", "author_url": "", "post_date": "2022-07-29T00:17:59.687000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1855758": "Hi there,\n\nYou can train offline and use the Kaggle notebook as just your inference code - however your training code may be requested in validating a model if you win.", "1855287": "You can upload the files you need as datasets, and with limitations (only the latest version is visible) notebook output - the recent cubert example (https://www.kaggle.com/code/nbroad/using-cubert/notebook) shows how to do this.\n\nFine-tuned models that you produce (and can reproduce) using your own code do *not* have to be public datasets.", "1854712": "I'm new on Kaggle too. However, I think I'm apt to try answer those two question. \n\nIn you kaggle notebook, you'll find an option in Settings that allow you to turn your internet on/off as shown in [this](https://www.kaggle.com/product-feedback/63544?adlt=strict&toWww=1&redig=F584FCFDC7C2404AB5E49409C9908D4E). \n\nThe runtime is for your entire notebook. Therefore, it should include training time too (though I'm not sure if it's possible to upload a trained model together the notebook).\n\n\n", "1853232": "Could someone clarify the rules? They state:\n```\nThis is a Code Competition\nSubmissions to this competition must be made through Notebooks. In order for the \"Submit\" button to be active after a commit, the following conditions must be met:\n\nCPU Notebook <= 9 hours run-time\nGPU Notebook <= 9 hours run-time\nInternet access disabled\nFreely & publicly available external data is allowed, including pre-trained models\n```\n\n1. If we need to submit via a notebook, and we can't access the internet, how do we upload our external data?\n\n2. Does 9 hours include training time? What if we train outside of the notebook, publish it freely, and then load it as \"free and public data\"?\n\nI'm new to Kaggle, so I don't know if these questions are unique to this challenge or apply to others.", "1875350": "" } }