Dear competition Host, We are currently working with one submission per day. This makes it very difficult for us to analyze the dataset and our code. Apart from that, some unnecessary errors are also made due to the small mistakes that we do unconsciously(Notebook Threw Exception,Notebook Exceeded Allowed Compute,Submission CSV Not Found,Submission Scoring Error). So we have to wait till the next day to submit the correct version. So please increase the submissions to at least 2 or 3.
", "messages": [ { "id": 1882320, "postDate": "2022-08-03T07:48:06.260Z", "content": "Dear competition Host, We are currently working with one submission per day. This makes it very difficult for us to analyze the dataset and our code. Apart from that, some unnecessary errors are also made due to the small mistakes that we do unconsciously(Notebook Threw Exception,Notebook Exceeded Allowed Compute,Submission CSV Not Found,Submission Scoring Error). So we have to wait till the next day to submit the correct version. So please increase the submissions to at least 2 or 3.
", "rawMarkdown": "Dear competition Host, We are currently working with one submission per day. This makes it very difficult for us to analyze the dataset and our code. Apart from that, some unnecessary errors are also made due to the small mistakes that we do unconsciously(**Notebook Threw Exception,Notebook Exceeded Allowed Compute,Submission CSV Not Found,Submission Scoring Error**). So we have to wait till the next day to submit the correct version. **So please increase the submissions to at least 2 or 3.**", "votes": 4 }, { "id": 1912213, "postDate": "2022-08-24T15:34:06.727Z", "content": "Getting successful submissions is always a bit of pain for me on these type of competitions. With only once chance per day I feel it would be a huge waste of time to do any work in this competition.
\nIt's always a huge shame when rules are created to prevent \"issues\" and they end up hurting the 99% of folks who don't cheat or don't probe.
", "rawMarkdown": "Getting successful submissions is always a bit of pain for me on these type of competitions. With only once chance per day I feel it would be a huge waste of time to do any work in this competition. \n\nIt's always a huge shame when rules are created to prevent \"issues\" and they end up hurting the 99% of folks who don't cheat or don't probe.", "votes": 1 }, { "id": 1882521, "postDate": "2022-08-03T10:16:45.770Z", "content": "They have already told us that they arent going to do that because of probing
", "rawMarkdown": "They have already told us that they arent going to do that because of probing", "votes": 1, "replies": [ { "id": 1882600, "postDate": "2022-08-03T10:56:23.050Z", "content": "ooooops😯😓
", "rawMarkdown": "ooooops😯😓" } ] }, { "id": 1952539, "postDate": "2022-09-23T18:25:55.233Z", "content": "9 hours run time is very little, too. The size of the test set : 93%/7% = 13.28, so the private tests will have 280*13,3=3724 images. Maybe I am wrong?
\nOne minute for one test would not be so bad in the real life.
I think 280 is the 100% private data size and the notebook runs on the entire dataset when you submit. Only 7% data is used to calculate scores for public LB. That's my understanding and I have some tests to hint this is so. I don't think kaggle will run all selected notebooks again on unseen hidden data and then release results, it would take weeks then. That cannot be the way, or am I too new to all this?
", "rawMarkdown": "I think 280 is the 100% private data size and the notebook runs on the entire dataset when you submit. Only 7% data is used to calculate scores for public LB. That's my understanding and I have some tests to hint this is so. I don't think kaggle will run all selected notebooks again on unseen hidden data and then release results, it would take weeks then. That cannot be the way, or am I too new to all this?", "votes": 1 }, { "id": 1971459, "postDate": "2022-10-04T16:16:20.990Z", "content": "Yes, you're correct. When you submit, they calculate both private and public scores, but only reveal public on the leaderboard until the end.
\nIf they re-ran them all, a lot of folks may end up getting surprised when their submits break at the end of the competition due to out of time or other unforseen errors.
", "rawMarkdown": "Yes, you're correct. When you submit, they calculate both private and public scores, but only reveal public on the leaderboard until the end. \n\nIf they re-ran them all, a lot of folks may end up getting surprised when their submits break at the end of the competition due to out of time or other unforseen errors.", "votes": 1 } ] }, { "id": 1952204, "postDate": "2022-09-23T13:54:08.707Z", "content": "…at least it could be for errors: just not counting failed submissions.
", "rawMarkdown": "...at least it could be for errors: just not counting failed submissions. ", "replies": [ { "id": 1971463, "postDate": "2022-10-04T16:18:14.483Z", "rawMarkdown": "", "isDeleted": true } ] } ], "comments": [ { "id": 1912213, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2022-08-24T15:34:06.727000", "content": "Getting successful submissions is always a bit of pain for me on these type of competitions. With only once chance per day I feel it would be a huge waste of time to do any work in this competition.
\nIt's always a huge shame when rules are created to prevent \"issues\" and they end up hurting the 99% of folks who don't cheat or don't probe.
", "votes": 1, "replies": [] }, { "id": 1882521, "author_name": "majoraregalia", "author_url": "", "post_date": "2022-08-03T10:16:45.770000", "content": "They have already told us that they arent going to do that because of probing
", "votes": 1, "replies": [ { "id": 1882600, "author_name": "nadhir hasan", "author_url": "", "post_date": "2022-08-03T10:56:23.050000", "content": "ooooops😯😓
", "votes": 0, "replies": [] } ] }, { "id": 1952539, "author_name": "Pierre Tisseur", "author_url": "", "post_date": "2022-09-23T18:25:55.233000", "content": "9 hours run time is very little, too. The size of the test set : 93%/7% = 13.28, so the private tests will have 280*13,3=3724 images. Maybe I am wrong?
\nOne minute for one test would not be so bad in the real life.
I think 280 is the 100% private data size and the notebook runs on the entire dataset when you submit. Only 7% data is used to calculate scores for public LB. That's my understanding and I have some tests to hint this is so. I don't think kaggle will run all selected notebooks again on unseen hidden data and then release results, it would take weeks then. That cannot be the way, or am I too new to all this?
", "votes": 1, "replies": [] }, { "id": 1971459, "author_name": "@kaggleqrdl", "author_url": "", "post_date": "2022-10-04T16:16:20.990000", "content": "Yes, you're correct. When you submit, they calculate both private and public scores, but only reveal public on the leaderboard until the end.
\nIf they re-ran them all, a lot of folks may end up getting surprised when their submits break at the end of the competition due to out of time or other unforseen errors.
", "votes": 1, "replies": [] } ] }, { "id": 1952204, "author_name": "MITEL-UNIUD", "author_url": "", "post_date": "2022-09-23T13:54:08.707000", "content": "…at least it could be for errors: just not counting failed submissions.
", "votes": 0, "replies": [ { "id": 1971463, "author_name": "", "author_url": "", "post_date": "2022-10-04T16:18:14.483000", "content": "", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "1882320": "Dear competition Host, We are currently working with one submission per day. This makes it very difficult for us to analyze the dataset and our code. Apart from that, some unnecessary errors are also made due to the small mistakes that we do unconsciously(**Notebook Threw Exception,Notebook Exceeded Allowed Compute,Submission CSV Not Found,Submission Scoring Error**). So we have to wait till the next day to submit the correct version. **So please increase the submissions to at least 2 or 3.**", "1912213": "Getting successful submissions is always a bit of pain for me on these type of competitions. With only once chance per day I feel it would be a huge waste of time to do any work in this competition. \n\nIt's always a huge shame when rules are created to prevent \"issues\" and they end up hurting the 99% of folks who don't cheat or don't probe.", "1882521": "They have already told us that they arent going to do that because of probing", "1952539": "9 hours run time is very little, too. The size of the test set : 93%/7% = 13.28, so the private tests will have 280*13,3=3724 images. Maybe I am wrong?\nOne minute for one test would not be so bad in the real life.", "1952204": "...at least it could be for errors: just not counting failed submissions. " } }