Here [1] the organizers say that \"Pre-trained models can be declared, as well; however, models resulting from your own original work, that you have trained yourself offline do not need to be shared/declared\", does this mean that in the final submission we can include several model checkpoints and run inference on them? Will this be ok if we create such checkpoints on a Kaggle TPU?
\n\n[1] https://www.kaggle.com/c/jigsaw-multilingual-toxic-comment-classification/discussion/138118
", "messages": [ { "id": 887531, "postDate": "2020-06-15T18:07:45.367Z", "content": "Yes, you can train offline or in kernels, and then do inference for submissions only. You can even only upload the csv file to a kernel and make a sub in this competition.
", "rawMarkdown": "Yes, you can train offline or in kernels, and then do inference for submissions only. You can even only upload the csv file to a kernel and make a sub in this competition. ", "votes": 3, "replies": [ { "id": 887595, "postDate": "2020-06-15T18:55:40Z", "content": "Thank you!
", "rawMarkdown": "Thank you!" } ] }, { "id": 887335, "postDate": "2020-06-15T16:03:32.403Z", "content": "Here [1] the organizers say that \"Pre-trained models can be declared, as well; however, models resulting from your own original work, that you have trained yourself offline do not need to be shared/declared\", does this mean that in the final submission we can include several model checkpoints and run inference on them? Will this be ok if we create such checkpoints on a Kaggle TPU?
\n\n[1] https://www.kaggle.com/c/jigsaw-multilingual-toxic-comment-classification/discussion/138118
", "rawMarkdown": "Here [1] the organizers say that \"Pre-trained models can be declared, as well; however, models resulting from your own original work, that you have trained yourself offline do not need to be shared/declared\", does this mean that in the final submission we can include several model checkpoints and run inference on them? Will this be ok if we create such checkpoints on a Kaggle TPU?\n\n[1] https://www.kaggle.com/c/jigsaw-multilingual-toxic-comment-classification/discussion/138118", "votes": 1 }, { "id": 892017, "postDate": "2020-06-18T15:42:59.270Z", "rawMarkdown": "", "votes": -1, "isDeleted": true } ], "comments": [ { "id": 887531, "author_name": "Psi", "author_url": "", "post_date": "2020-06-15T18:07:45.367000", "content": "Yes, you can train offline or in kernels, and then do inference for submissions only. You can even only upload the csv file to a kernel and make a sub in this competition.
", "votes": 3, "replies": [ { "id": 887595, "author_name": "arseny-n", "author_url": "", "post_date": "2020-06-15T18:55:40", "content": "Thank you!
", "votes": 0, "replies": [] } ] }, { "id": 892017, "author_name": "", "author_url": "", "post_date": "2020-06-18T15:42:59.270000", "content": "", "votes": -1, "replies": [] } ], "raw_markdown_by_id": { "887531": "Yes, you can train offline or in kernels, and then do inference for submissions only. You can even only upload the csv file to a kernel and make a sub in this competition. ", "887335": "Here [1] the organizers say that \"Pre-trained models can be declared, as well; however, models resulting from your own original work, that you have trained yourself offline do not need to be shared/declared\", does this mean that in the final submission we can include several model checkpoints and run inference on them? Will this be ok if we create such checkpoints on a Kaggle TPU?\n\n[1] https://www.kaggle.com/c/jigsaw-multilingual-toxic-comment-classification/discussion/138118", "892017": "" } }