Unless your team manually selects two preferred submissions, our system will select submissions for you based on the highest public score your team has achieved.
\n\nYou probably do not want to accidentally use any of the popular \"quick check\" or \"fast submission\" style kernels for your top two.
", "messages": [ { "id": "619137", "postDate": "09/05/2019 21:18:29", "content": "Unless your team manually selects two preferred submissions, our system will select submissions for you based on the highest public score your team has achieved.
\n\nYou probably do not want to accidentally use any of the popular \"quick check\" or \"fast submission\" style kernels for your top two.
", "rawMarkdown": "Unless your team manually selects two preferred submissions, our system will select submissions for you based on the highest public score your team has achieved.\n\nYou probably do not want to accidentally use any of the popular \"quick check\" or \"fast submission\" style kernels for your top two.", "votes": null }, { "id": "619245", "postDate": "09/06/2019 01:34:55", "content": "@sohier For the final private lb, will the mean or maximum score of the two submissions be selected?
", "rawMarkdown": "sohier For the final private lb, will the mean or maximum score of the two submissions be selected?", "votes": null }, { "id": "619277", "postDate": "09/06/2019 02:54:56", "content": "Also, can we use kernel option internte on for final submission? It's for mixed precision training(now need pytorch 1.1.0)
", "rawMarkdown": "Also, can we use kernel option internte on for final submission? It's for mixed precision training(now need pytorch 1.1.0)", "votes": null }, { "id": "619323", "postDate": "09/06/2019 04:38:45", "content": "the highest one will be selected
", "rawMarkdown": "the highest one will be selected", "votes": null }, { "id": "619324", "postDate": "09/06/2019 04:41:31", "content": "no, it should be turned off
", "rawMarkdown": "no, it should be turned off", "votes": null }, { "id": "619545", "postDate": "09/06/2019 09:17:23", "content": "Ok, so there's no way to use mixed precision? @nuller
", "rawMarkdown": "Ok, so there's no way to use mixed precision? @nuller", "votes": null }, { "id": "619571", "postDate": "09/06/2019 09:50:51", "content": "@homoalways see this kernel https://www.kaggle.com/chanhu/eye-efficientnet-pytorch-lb-0-777
", "rawMarkdown": "homoalways see this kernel https://www.kaggle.com/chanhu/eye-efficientnet-pytorch-lb-0-777", "votes": null }, { "id": "620992", "postDate": "09/08/2019 06:13:18", "content": "Dear the competition organizers,
\n\nThanks for your hard work in organizing the competition APTOS 2019 Blindness Detection！
\n\nBut we were ranked 192, which is obviously very disappointing because we were ranked 52 before the final submission. According to the competition rules, two models that performed best on public test data were submitted and they were run on private test data. Then, our ranking went down from 52 to 192 - down by 140!
\n\nWe do not have the right to challenge the rules. However, such rules encourage random ranking instead of hard work. This can be seen in other teams, for example, the team ranked 24 went straight down to 291, and the team ranked 2nd went down to 83!
\n\nWhat's more, Kaggle has already run all submitted models on private data, and our best model on private data can achieve 0.922 (currently 0.918). Then, why didnot Kaggle use the best model?
\n\nHope Kaggle and the organizers can review the rules, and we look forward to taking more competitions in future.
", "rawMarkdown": "Dear the competition organizers, \n\nThanks for your hard work in organizing the competition APTOS 2019 Blindness Detection！ \n\nBut we were ranked 192, which is obviously very disappointing because we were ranked 52 before the final submission. According to the competition rules, two models that performed best on public test data were submitted and they were run on private test data. Then, our ranking went down from 52 to 192 - down by 140! \n\nWe do not have the right to challenge the rules. However, such rules encourage random ranking instead of hard work. This can be seen in other teams, for example, the team ranked 24 went straight down to 291, and the team ranked 2nd went down to 83!\n\nWhat's more, Kaggle has already run all submitted models on private data, and our best model on private data can achieve 0.922 (currently 0.918). Then, why didnot Kaggle use the best model? \n\nHope Kaggle and the organizers can review the rules, and we look forward to taking more competitions in future. ", "votes": null }, { "id": "621516", "postDate": "09/08/2019 16:00:16", "content": "I am happy to participate in this competition for the first time at Kaggle. I learned a lot, I hope to write public kernels next time to share the improvements I got. Mostly I started reading the discussions at the end of the competition and realized that it helps a lot. I will read much more next time. Unfortunately I left my kernels with fast submission, so I got a private score of 0.0000, but I'm glad one of my Kernels got a private score of 88%. So I leave the tip that it is very important to read the discussions, if I had read here, would see that could not send quick submissions. Thank you all!
", "rawMarkdown": "I am happy to participate in this competition for the first time at Kaggle. I learned a lot, I hope to write public kernels next time to share the improvements I got. Mostly I started reading the discussions at the end of the competition and realized that it helps a lot. I will read much more next time. Unfortunately I left my kernels with fast submission, so I got a private score of 0.0000, but I'm glad one of my Kernels got a private score of 88%. So I leave the tip that it is very important to read the discussions, if I had read here, would see that could not send quick submissions. Thank you all!", "votes": null } ], "comments": [ { "id": 619245, "author_name": "homoalways", "author_url": "", "post_date": "09/06/2019 01:34:55", "content": "@sohier For the final private lb, will the mean or maximum score of the two submissions be selected?
", "votes": null, "replies": [ { "id": 619323, "author_name": "nuller", "author_url": "", "post_date": "09/06/2019 04:38:45", "content": "the highest one will be selected
", "votes": null, "replies": [] } ] }, { "id": 619277, "author_name": "homoalways", "author_url": "", "post_date": "09/06/2019 02:54:56", "content": "Also, can we use kernel option internte on for final submission? It's for mixed precision training(now need pytorch 1.1.0)
", "votes": null, "replies": [ { "id": 619324, "author_name": "nuller", "author_url": "", "post_date": "09/06/2019 04:41:31", "content": "no, it should be turned off
", "votes": null, "replies": [] }, { "id": 619545, "author_name": "homoalways", "author_url": "", "post_date": "09/06/2019 09:17:23", "content": "Ok, so there's no way to use mixed precision? @nuller
", "votes": null, "replies": [] }, { "id": 619571, "author_name": "nuller", "author_url": "", "post_date": "09/06/2019 09:50:51", "content": "@homoalways see this kernel https://www.kaggle.com/chanhu/eye-efficientnet-pytorch-lb-0-777
", "votes": null, "replies": [] } ] }, { "id": 620992, "author_name": "lifengnan", "author_url": "", "post_date": "09/08/2019 06:13:18", "content": "Dear the competition organizers,
\n\nThanks for your hard work in organizing the competition APTOS 2019 Blindness Detection！
\n\nBut we were ranked 192, which is obviously very disappointing because we were ranked 52 before the final submission. According to the competition rules, two models that performed best on public test data were submitted and they were run on private test data. Then, our ranking went down from 52 to 192 - down by 140!
\n\nWe do not have the right to challenge the rules. However, such rules encourage random ranking instead of hard work. This can be seen in other teams, for example, the team ranked 24 went straight down to 291, and the team ranked 2nd went down to 83!
\n\nWhat's more, Kaggle has already run all submitted models on private data, and our best model on private data can achieve 0.922 (currently 0.918). Then, why didnot Kaggle use the best model?
\n\nHope Kaggle and the organizers can review the rules, and we look forward to taking more competitions in future.
", "votes": null, "replies": [] }, { "id": 621516, "author_name": "custodiogabriel", "author_url": "", "post_date": "09/08/2019 16:00:16", "content": "I am happy to participate in this competition for the first time at Kaggle. I learned a lot, I hope to write public kernels next time to share the improvements I got. Mostly I started reading the discussions at the end of the competition and realized that it helps a lot. I will read much more next time. Unfortunately I left my kernels with fast submission, so I got a private score of 0.0000, but I'm glad one of my Kernels got a private score of 88%. So I leave the tip that it is very important to read the discussions, if I had read here, would see that could not send quick submissions. Thank you all!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "619137": "Unless your team manually selects two preferred submissions, our system will select submissions for you based on the highest public score your team has achieved.\n\nYou probably do not want to accidentally use any of the popular \"quick check\" or \"fast submission\" style kernels for your top two.", "619245": "sohier For the final private lb, will the mean or maximum score of the two submissions be selected?", "619277": "Also, can we use kernel option internte on for final submission? It's for mixed precision training(now need pytorch 1.1.0)", "619323": "the highest one will be selected", "619324": "no, it should be turned off", "619545": "Ok, so there's no way to use mixed precision? @nuller", "619571": "homoalways see this kernel https://www.kaggle.com/chanhu/eye-efficientnet-pytorch-lb-0-777", "620992": "Dear the competition organizers, \n\nThanks for your hard work in organizing the competition APTOS 2019 Blindness Detection！ \n\nBut we were ranked 192, which is obviously very disappointing because we were ranked 52 before the final submission. According to the competition rules, two models that performed best on public test data were submitted and they were run on private test data. Then, our ranking went down from 52 to 192 - down by 140! \n\nWe do not have the right to challenge the rules. However, such rules encourage random ranking instead of hard work. This can be seen in other teams, for example, the team ranked 24 went straight down to 291, and the team ranked 2nd went down to 83!\n\nWhat's more, Kaggle has already run all submitted models on private data, and our best model on private data can achieve 0.922 (currently 0.918). Then, why didnot Kaggle use the best model? \n\nHope Kaggle and the organizers can review the rules, and we look forward to taking more competitions in future. ", "621516": "I am happy to participate in this competition for the first time at Kaggle. I learned a lot, I hope to write public kernels next time to share the improvements I got. Mostly I started reading the discussions at the end of the competition and realized that it helps a lot. I will read much more next time. Unfortunately I left my kernels with fast submission, so I got a private score of 0.0000, but I'm glad one of my Kernels got a private score of 88%. So I leave the tip that it is very important to read the discussions, if I had read here, would see that could not send quick submissions. Thank you all!" }, "source": "meta" }