In the Overview, under Submission Files it says \"For the tile:xla collection, only the first 5 entries will be considered and the rest will be ignored.
\nIs this accurate or is it supposed to be only the last entries will be considered?
", "messages": [ { "id": 2414885, "postDate": "2023-08-30T03:11:16.257Z", "content": "Ahh … I was confusing \"entries\" with \"submissions.\" Thank you for all of this helpful information.
", "rawMarkdown": "Ahh ... I was confusing \"entries\" with \"submissions.\" Thank you for all of this helpful information.", "votes": 1 }, { "id": 2414738, "postDate": "2023-08-29T23:47:35.243Z", "content": "It is accurate.
\nIt is a ranking problem seeking the fastest (i.e., least) runtimes. Your model should rank fast-to-slow i.e. sort items in (default) ascending order. The starter code is setup that way. We take the first 5 entries (which should correspond to the 5 smallest runtimes, predicted by your model)
", "rawMarkdown": "It is accurate.\n\nIt is a ranking problem seeking the **fastest** (i.e., **least**) runtimes. Your model should rank fast-to-slow i.e. sort items in (default) ascending order. The starter code is setup that way. We take the first 5 entries (which should correspond to the 5 smallest runtimes, predicted by your model)", "votes": 1, "replies": [ { "id": 2414877, "postDate": "2023-08-30T02:48:10.870Z", "content": "Hi, can we develop the model in pytorch or are TF models a strict requirement. Couldn't see any mention of it but I think it might be critical to subsequent integration to XLA
", "rawMarkdown": "Hi, can we develop the model in pytorch or are TF models a strict requirement. Couldn't see any mention of it but I think it might be critical to subsequent integration to XLA", "replies": [ { "id": 2414998, "postDate": "2023-08-30T04:02:32.977Z", "content": "You can development your model in any framework of your choice. You only need to submit the result in the right format. In fact, one of the existing layout model baselines is implemented in PyTorch: https://github.com/kaidic/GST.
", "rawMarkdown": "You can development your model in any framework of your choice. You only need to submit the result in the right format. In fact, one of the existing layout model baselines is implemented in PyTorch: https://github.com/kaidic/GST." } ] }, { "id": 2414971, "postDate": "2023-08-30T03:53:15.830Z", "content": "Thank you for this clarifying information.
", "rawMarkdown": "Thank you for this clarifying information." } ] }, { "id": 2414734, "postDate": "2023-08-29T23:38:14.890Z", "content": "In the Overview, under Submission Files it says \"For the tile:xla collection, only the first 5 entries will be considered and the rest will be ignored.
\nIs this accurate or is it supposed to be only the last entries will be considered?
", "rawMarkdown": "In the Overview, under Submission Files it says \"For the tile:xla collection, only the first 5 entries will be considered and the rest will be ignored.\n\nIs this accurate or is it supposed to be only the last entries will be considered?", "votes": 2 }, { "id": 2414739, "postDate": "2023-08-29T23:48:01.057Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 2414885, "author_name": "Sarah Hardy", "author_url": "", "post_date": "2023-08-30T03:11:16.257000", "content": "Ahh … I was confusing \"entries\" with \"submissions.\" Thank you for all of this helpful information.
", "votes": 1, "replies": [] }, { "id": 2414738, "author_name": "Sami Abu-El-Haija", "author_url": "", "post_date": "2023-08-29T23:47:35.243000", "content": "It is accurate.
\nIt is a ranking problem seeking the fastest (i.e., least) runtimes. Your model should rank fast-to-slow i.e. sort items in (default) ascending order. The starter code is setup that way. We take the first 5 entries (which should correspond to the 5 smallest runtimes, predicted by your model)
", "votes": 1, "replies": [ { "id": 2414877, "author_name": "Lakshmi Ramesh", "author_url": "", "post_date": "2023-08-30T02:48:10.870000", "content": "Hi, can we develop the model in pytorch or are TF models a strict requirement. Couldn't see any mention of it but I think it might be critical to subsequent integration to XLA
", "votes": 0, "replies": [ { "id": 2414998, "author_name": "Mangpo Phothilimthana", "author_url": "", "post_date": "2023-08-30T04:02:32.977000", "content": "You can development your model in any framework of your choice. You only need to submit the result in the right format. In fact, one of the existing layout model baselines is implemented in PyTorch: https://github.com/kaidic/GST.
", "votes": 0, "replies": [] } ] }, { "id": 2414971, "author_name": "Sarah Hardy", "author_url": "", "post_date": "2023-08-30T03:53:15.830000", "content": "Thank you for this clarifying information.
", "votes": 0, "replies": [] } ] }, { "id": 2414739, "author_name": "", "author_url": "", "post_date": "2023-08-29T23:48:01.057000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "2414885": "Ahh ... I was confusing \"entries\" with \"submissions.\" Thank you for all of this helpful information.", "2414738": "It is accurate.\n\nIt is a ranking problem seeking the **fastest** (i.e., **least**) runtimes. Your model should rank fast-to-slow i.e. sort items in (default) ascending order. The starter code is setup that way. We take the first 5 entries (which should correspond to the 5 smallest runtimes, predicted by your model)", "2414734": "In the Overview, under Submission Files it says \"For the tile:xla collection, only the first 5 entries will be considered and the rest will be ignored.\n\nIs this accurate or is it supposed to be only the last entries will be considered?", "2414739": "" } }