In other similar competitions we have at least 5 submissions allowed per day. Having only 2 submissions restricts us in terms of understanding which model works best. Is it feasible to expand the allowable submission limit per day?
", "messages": [ { "id": "1262317", "postDate": "04/04/2021 05:44:04", "content": "In other similar competitions we have at least 5 submissions allowed per day. Having only 2 submissions restricts us in terms of understanding which model works best. Is it feasible to expand the allowable submission limit per day?
", "rawMarkdown": "In other similar competitions we have at least 5 submissions allowed per day. Having only 2 submissions restricts us in terms of understanding which model works best. Is it feasible to expand the allowable submission limit per day?", "votes": null }, { "id": "1262506", "postDate": "04/04/2021 11:46:46", "content": "code + only 2 subs a day -> LB probing is not a low hanging fruit.
", "rawMarkdown": "code + only 2 subs a day -> LB probing is not a low hanging fruit.", "votes": null }, { "id": "1263055", "postDate": "04/05/2021 04:08:52", "content": "I am pretty new to competitions, but I think limiting the submissions helps CPU users since the maximum runtime for CPU notebooks is more than double the GPU runtime. If more than 3 submissions were allowed per day GPU users would have a significant advantage in the number of submissions.
\nI am curious to know what a CPU only solution would look like since almost everyone in literature uses a CNN on the spectrograms. (Or a close cousin of that.)
", "rawMarkdown": "I am pretty new to competitions, but I think limiting the submissions helps CPU users since the maximum runtime for CPU notebooks is more than double the GPU runtime. If more than 3 submissions were allowed per day GPU users would have a significant advantage in the number of submissions.\n\nI am curious to know what a CPU only solution would look like since almost everyone in literature uses a CNN on the spectrograms. (Or a close cousin of that.)", "votes": null }, { "id": "1263144", "postDate": "04/05/2021 05:55:11", "content": "Spot on @cpmpml !
", "rawMarkdown": "Spot on @cpmpml !", "votes": null }, { "id": "1265343", "postDate": "04/06/2021 20:12:23", "content": "Hi all! Sorry for the silly question but I don´t know how to use the test set to make submission.csv, how can I see test data? thanks in advance!
", "rawMarkdown": "Hi all! Sorry for the silly question but I don´t know how to use the test set to make submission.csv, how can I see test data? thanks in advance!", "votes": null }, { "id": "1265432", "postDate": "04/06/2021 22:03:34", "content": "You have to write code that will read test data and predict on it. Look for the Code tab, it contains notebooks that do it, for instance this one: https://www.kaggle.com/hidehisaarai1213/pytorch-inference-birdclef2021-starter
", "rawMarkdown": "You have to write code that will read test data and predict on it. Look for the Code tab, it contains notebooks that do it, for instance this one: https://www.kaggle.com/hidehisaarai1213/pytorch-inference-birdclef2021-starter", "votes": null }, { "id": "1265464", "postDate": "04/06/2021 23:08:17", "content": "Thank you for your kind advise! I though there should be and csv file with audio names, but I guess that should be looked on the folder for the given ogg files.
", "rawMarkdown": "Thank you for your kind advise! I though there should be and csv file with audio names, but I guess that should be looked on the folder for the given ogg files.", "votes": null } ], "comments": [ { "id": 1262506, "author_name": "cpmpml", "author_url": "", "post_date": "04/04/2021 11:46:46", "content": "code + only 2 subs a day -> LB probing is not a low hanging fruit.
", "votes": null, "replies": [ { "id": 1263144, "author_name": "saurabhbagchi", "author_url": "", "post_date": "04/05/2021 05:55:11", "content": "Spot on @cpmpml !
", "votes": null, "replies": [] } ] }, { "id": 1263055, "author_name": "aarons0", "author_url": "", "post_date": "04/05/2021 04:08:52", "content": "I am pretty new to competitions, but I think limiting the submissions helps CPU users since the maximum runtime for CPU notebooks is more than double the GPU runtime. If more than 3 submissions were allowed per day GPU users would have a significant advantage in the number of submissions.
\nI am curious to know what a CPU only solution would look like since almost everyone in literature uses a CNN on the spectrograms. (Or a close cousin of that.)
", "votes": null, "replies": [] }, { "id": 1265343, "author_name": "andrescastroitm", "author_url": "", "post_date": "04/06/2021 20:12:23", "content": "Hi all! Sorry for the silly question but I don´t know how to use the test set to make submission.csv, how can I see test data? thanks in advance!
", "votes": null, "replies": [ { "id": 1265432, "author_name": "cpmpml", "author_url": "", "post_date": "04/06/2021 22:03:34", "content": "You have to write code that will read test data and predict on it. Look for the Code tab, it contains notebooks that do it, for instance this one: https://www.kaggle.com/hidehisaarai1213/pytorch-inference-birdclef2021-starter
", "votes": null, "replies": [] }, { "id": 1265464, "author_name": "andrescastroitm", "author_url": "", "post_date": "04/06/2021 23:08:17", "content": "Thank you for your kind advise! I though there should be and csv file with audio names, but I guess that should be looked on the folder for the given ogg files.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1262317": "In other similar competitions we have at least 5 submissions allowed per day. Having only 2 submissions restricts us in terms of understanding which model works best. Is it feasible to expand the allowable submission limit per day?", "1262506": "code + only 2 subs a day -> LB probing is not a low hanging fruit.", "1263055": "I am pretty new to competitions, but I think limiting the submissions helps CPU users since the maximum runtime for CPU notebooks is more than double the GPU runtime. If more than 3 submissions were allowed per day GPU users would have a significant advantage in the number of submissions.\n\nI am curious to know what a CPU only solution would look like since almost everyone in literature uses a CNN on the spectrograms. (Or a close cousin of that.)", "1263144": "Spot on @cpmpml !", "1265343": "Hi all! Sorry for the silly question but I don´t know how to use the test set to make submission.csv, how can I see test data? thanks in advance!", "1265432": "You have to write code that will read test data and predict on it. Look for the Code tab, it contains notebooks that do it, for instance this one: https://www.kaggle.com/hidehisaarai1213/pytorch-inference-birdclef2021-starter", "1265464": "Thank you for your kind advise! I though there should be and csv file with audio names, but I guess that should be looked on the folder for the given ogg files." }, "source": "meta" }