Hello,
\n\nam I right with my understanding off the rules?
\n\nYou have to write a notebook, that calculates a submission in 9 hours or less. You must run it, on one of kaggles P100 GPUs. No private external data is allowed. That means, you can't train your own private model at home and use it in the notebook.
\n\nI wonder, why the dataset is about 472GB. How can use process this amount of data in this short time?
", "messages": [ { "id": 700658, "postDate": "2019-12-22T11:29:00.023Z", "content": "Hello,
\n\nam I right with my understanding off the rules?
\n\nYou have to write a notebook, that calculates a submission in 9 hours or less. You must run it, on one of kaggles P100 GPUs. No private external data is allowed. That means, you can't train your own private model at home and use it in the notebook.
\n\nI wonder, why the dataset is about 472GB. How can use process this amount of data in this short time?
", "rawMarkdown": "Hello,\n\nam I right with my understanding off the rules?\n\nYou have to write a notebook, that calculates a submission in 9 hours or less. You must run it, on one of kaggles P100 GPUs. No private external data is allowed. That means, you can't train your own private model at home and use it in the notebook.\n\nI wonder, why the dataset is about 472GB. How can use process this amount of data in this short time?", "votes": 1 }, { "id": 701045, "postDate": "2019-12-23T02:50:03.247Z", "content": "It might be helpful for you to read Getting Started :)
\n\nhttps://www.kaggle.com/c/deepfake-detection-challenge/overview/getting-started
", "rawMarkdown": "It might be helpful for you to read Getting Started :) \n\nhttps://www.kaggle.com/c/deepfake-detection-challenge/overview/getting-started" }, { "id": 700702, "postDate": "2019-12-22T13:22:32.110Z", "content": "\n\n\nThat means, you can't train your own private model at home and use it in the notebook.
\n
That is incorrect. In fact, you are supposed to train your own private model at home and use it in the notebook.
\n\nThe \"9 hours or less\" limit is for making predictions using your already-trained model on the public test set of 4000 videos.
", "rawMarkdown": "> That means, you can't train your own private model at home and use it in the notebook.\n\nThat is incorrect. In fact, you are *supposed* to train your own private model at home and use it in the notebook.\n\nThe \"9 hours or less\" limit is for making predictions using your already-trained model on the public test set of 4000 videos.", "replies": [ { "id": 700896, "postDate": "2019-12-22T19:33:16.150Z", "content": "Than I missinterpret the following rule: \"External data is allowed up to 1 GB in size. External data must be freely & publicly available, including pre-trained models.\"
\n\nI can upload my modell with a maximum size of 1 GB, but it must be publicly available, not?
\n\nWhere is the part in the rules, that allows me to upload my private model?
", "rawMarkdown": "Than I missinterpret the following rule: \"External data is allowed up to 1 GB in size. External data must be freely & publicly available, including pre-trained models.\"\n\nI can upload my modell with a maximum size of 1 GB, but it must be publicly available, not?\n\nWhere is the part in the rules, that allows me to upload my private model?" }, { "id": 700930, "postDate": "2019-12-22T20:41:47.913Z", "content": "See this comment by the competition organizer: https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121203#693874
", "rawMarkdown": "See this comment by the competition organizer: https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121203#693874" }, { "id": 701000, "postDate": "2019-12-23T00:34:35.930Z", "content": "\"I can upload my modell with a maximum size of 1 GB, but it must be publicly available, not?\"
\n\nNo - YOUR model does not have to be available to public.
", "rawMarkdown": "\"I can upload my modell with a maximum size of 1 GB, but it must be publicly available, not?\"\n\nNo - YOUR model does not have to be available to public. ", "votes": 1 }, { "id": 701729, "postDate": "2019-12-23T20:55:40.693Z", "content": "@kinimodsreve When you write your code in Kaggle's notebook/script interface, you can upload a dataset as external data that feeds into the notebook. That external data source can be a model trained offline and cannot exceed 1 GB. External data (including external trained models) are permitted in the rule.s The external trained model does not need to be shared publicly, but if you're using any pre-trained models or external datasets to train your model, then you do need to disclose those publicly.
", "rawMarkdown": "@kinimodsreve When you write your code in Kaggle's notebook/script interface, you can upload a dataset as external data that feeds into the notebook. That external data source can be a model trained offline and cannot exceed 1 GB. External data (including external trained models) are permitted in the rule.s The external trained model does not need to be shared publicly, but if you're using any pre-trained models or external datasets to train your model, then you do need to disclose those publicly. ", "votes": 1 }, { "id": 702948, "postDate": "2019-12-25T11:45:41.177Z", "content": "Thanks to all for clarification!
", "rawMarkdown": "Thanks to all for clarification!" } ] }, { "id": 700861, "postDate": "2019-12-22T18:31:34.597Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 701045, "author_name": "hirviö", "author_url": "", "post_date": "2019-12-23T02:50:03.247000", "content": "It might be helpful for you to read Getting Started :)
\n\nhttps://www.kaggle.com/c/deepfake-detection-challenge/overview/getting-started
", "votes": 0, "replies": [] }, { "id": 700702, "author_name": "Human Analog", "author_url": "", "post_date": "2019-12-22T13:22:32.110000", "content": "\n\n\nThat means, you can't train your own private model at home and use it in the notebook.
\n
That is incorrect. In fact, you are supposed to train your own private model at home and use it in the notebook.
\n\nThe \"9 hours or less\" limit is for making predictions using your already-trained model on the public test set of 4000 videos.
", "votes": 0, "replies": [ { "id": 700896, "author_name": "Kinimod", "author_url": "", "post_date": "2019-12-22T19:33:16.150000", "content": "Than I missinterpret the following rule: \"External data is allowed up to 1 GB in size. External data must be freely & publicly available, including pre-trained models.\"
\n\nI can upload my modell with a maximum size of 1 GB, but it must be publicly available, not?
\n\nWhere is the part in the rules, that allows me to upload my private model?
", "votes": 0, "replies": [] }, { "id": 700930, "author_name": "Human Analog", "author_url": "", "post_date": "2019-12-22T20:41:47.913000", "content": "See this comment by the competition organizer: https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121203#693874
", "votes": 0, "replies": [] }, { "id": 701000, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2019-12-23T00:34:35.930000", "content": "\"I can upload my modell with a maximum size of 1 GB, but it must be publicly available, not?\"
\n\nNo - YOUR model does not have to be available to public.
", "votes": 1, "replies": [] }, { "id": 701729, "author_name": "Julia Elliott", "author_url": "", "post_date": "2019-12-23T20:55:40.693000", "content": "@kinimodsreve When you write your code in Kaggle's notebook/script interface, you can upload a dataset as external data that feeds into the notebook. That external data source can be a model trained offline and cannot exceed 1 GB. External data (including external trained models) are permitted in the rule.s The external trained model does not need to be shared publicly, but if you're using any pre-trained models or external datasets to train your model, then you do need to disclose those publicly.
", "votes": 1, "replies": [] }, { "id": 702948, "author_name": "Kinimod", "author_url": "", "post_date": "2019-12-25T11:45:41.177000", "content": "Thanks to all for clarification!
", "votes": 0, "replies": [] } ] }, { "id": 700861, "author_name": "", "author_url": "", "post_date": "2019-12-22T18:31:34.597000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "700658": "Hello,\n\nam I right with my understanding off the rules?\n\nYou have to write a notebook, that calculates a submission in 9 hours or less. You must run it, on one of kaggles P100 GPUs. No private external data is allowed. That means, you can't train your own private model at home and use it in the notebook.\n\nI wonder, why the dataset is about 472GB. How can use process this amount of data in this short time?", "701045": "It might be helpful for you to read Getting Started :) \n\nhttps://www.kaggle.com/c/deepfake-detection-challenge/overview/getting-started", "700702": "> That means, you can't train your own private model at home and use it in the notebook.\n\nThat is incorrect. In fact, you are *supposed* to train your own private model at home and use it in the notebook.\n\nThe \"9 hours or less\" limit is for making predictions using your already-trained model on the public test set of 4000 videos.", "700861": "" } }