Hey so this is a featured code competition so you can only submit when you commit your notebook ? if so how are you guys managing to train your models ? is it through using kaggle notebooks ? I'm using GPU with pytorch and an epoch takes about 40-50 min. I haven't tried TPU's but the size of data tells me it's not that fast aswell. Are you guys training on your computers or VM and then uploading your weights or what are doing exactly ? Thanks
", "messages": [ { "id": 895780, "postDate": "2020-06-21T15:48:21.993Z", "content": "You can train offline and upload your weights as a dataset that you then add to your submission notebook. For predicting/submitting then yes everything must happen in a kernel.
", "rawMarkdown": "You can train offline and upload your weights as a dataset that you then add to your submission notebook. For predicting/submitting then yes everything must happen in a kernel.", "votes": 1, "replies": [ { "id": 895795, "postDate": "2020-06-21T16:02:09.427Z", "rawMarkdown": "", "isDeleted": true } ] }, { "id": 895230, "postDate": "2020-06-21T07:32:24.447Z", "content": "Hey so this is a featured code competition so you can only submit when you commit your notebook ? if so how are you guys managing to train your models ? is it through using kaggle notebooks ? I'm using GPU with pytorch and an epoch takes about 40-50 min. I haven't tried TPU's but the size of data tells me it's not that fast aswell. Are you guys training on your computers or VM and then uploading your weights or what are doing exactly ? Thanks
", "rawMarkdown": "Hey so this is a featured code competition so you can only submit when you commit your notebook ? if so how are you guys managing to train your models ? is it through using kaggle notebooks ? I'm using GPU with pytorch and an epoch takes about 40-50 min. I haven't tried TPU's but the size of data tells me it's not that fast aswell. Are you guys training on your computers or VM and then uploading your weights or what are doing exactly ? Thanks ", "votes": 1 }, { "id": 920368, "postDate": "2020-07-08T14:47:10.907Z", "content": "It is pretty clear that we are training our own models and we cna upload them, good, but the submission process is still unlcear, even after following all the kaggle instructions, the submitted file is still 3 lines, is there any more detailed explanation on the submission process?
", "rawMarkdown": "It is pretty clear that we are training our own models and we cna upload them, good, but the submission process is still unlcear, even after following all the kaggle instructions, the submitted file is still 3 lines, is there any more detailed explanation on the submission process?", "replies": [ { "id": 928337, "postDate": "2020-07-13T22:24:44.013Z", "rawMarkdown": "", "isDeleted": true } ] }, { "id": 905406, "postDate": "2020-06-28T14:29:09.210Z", "content": "You can train your model offline or on VM then upload to the notebook, and use the notebook to produce the output
", "rawMarkdown": "You can train your model offline or on VM then upload to the notebook, and use the notebook to produce the output" }, { "id": 896178, "postDate": "2020-06-21T23:47:15.490Z", "content": "I am using Kaggle Notebooks. xD
", "rawMarkdown": "I am using Kaggle Notebooks. xD" } ], "comments": [ { "id": 895780, "author_name": "Arnaud Roussel", "author_url": "", "post_date": "2020-06-21T15:48:21.993000", "content": "You can train offline and upload your weights as a dataset that you then add to your submission notebook. For predicting/submitting then yes everything must happen in a kernel.
", "votes": 1, "replies": [ { "id": 895795, "author_name": "", "author_url": "", "post_date": "2020-06-21T16:02:09.427000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 920368, "author_name": "Leslie Solorzano", "author_url": "", "post_date": "2020-07-08T14:47:10.907000", "content": "It is pretty clear that we are training our own models and we cna upload them, good, but the submission process is still unlcear, even after following all the kaggle instructions, the submitted file is still 3 lines, is there any more detailed explanation on the submission process?
", "votes": 0, "replies": [ { "id": 928337, "author_name": "", "author_url": "", "post_date": "2020-07-13T22:24:44.013000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 905406, "author_name": "Deemo Chen", "author_url": "", "post_date": "2020-06-28T14:29:09.210000", "content": "You can train your model offline or on VM then upload to the notebook, and use the notebook to produce the output
", "votes": 0, "replies": [] }, { "id": 896178, "author_name": "Salman", "author_url": "", "post_date": "2020-06-21T23:47:15.490000", "content": "I am using Kaggle Notebooks. xD
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "895780": "You can train offline and upload your weights as a dataset that you then add to your submission notebook. For predicting/submitting then yes everything must happen in a kernel.", "895230": "Hey so this is a featured code competition so you can only submit when you commit your notebook ? if so how are you guys managing to train your models ? is it through using kaggle notebooks ? I'm using GPU with pytorch and an epoch takes about 40-50 min. I haven't tried TPU's but the size of data tells me it's not that fast aswell. Are you guys training on your computers or VM and then uploading your weights or what are doing exactly ? Thanks ", "920368": "It is pretty clear that we are training our own models and we cna upload them, good, but the submission process is still unlcear, even after following all the kaggle instructions, the submitted file is still 3 lines, is there any more detailed explanation on the submission process?", "905406": "You can train your model offline or on VM then upload to the notebook, and use the notebook to produce the output", "896178": "I am using Kaggle Notebooks. xD" } }