0.96+ on LB) and submission.csv from best public kernel(0.9663+) and compared the target values. This is what I got \n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1528571%2F60a56a9f688ba2c9b164239a4868bb8c%2Flb_manual.png?generation=1578124759033508&alt=media\")
I took my 4 best models(0.96+ on LB) and submission.csv from best public kernel(0.9663+) and compared the target values. This is what I got \n
They are pretty consistent excpet for one: Test_6_grapheme_root.
Does this mean that models with 0.97+ on LB will predict different values for x number of the row_id? What could be the possible x?
I took my 4 best models(0.96+ on LB) and submission.csv from best public kernel(0.9663+) and compared the target values. This is what I got \n
They are pretty consistent excpet for one: Test_6_grapheme_root.
Does this mean that models with 0.97+ on LB will predict different values for x number of the row_id? What could be the possible x?
I don't think you can look at the sample submission to infer anything with respect to differences in LB scores since 12 test files is probably just enough to make sure your notebook runs correctly. It gets re-run on full test set once you hit submit.
\n\n\n", "rawMarkdown": "I don't think you can look at the sample submission to infer anything with respect to differences in LB scores since 12 test files is probably just enough to make sure your notebook runs correctly. It gets re-run on full test set once you hit submit.\n\n> This is a synchronous rerun code competition, you can assume that the complete test set will contain essentially the same size and number of images as the training set. Consider performing inference on just one batch at a time to avoid memory errors. Only the first few rows/images in the test set and sample submission files can be downloaded. These samples provided so you can review the basic structure of the files and to ensure consistency between the publicly available set of file names and those your code will have access to while it is being rerun for scoring.", "votes": 3 }, { "id": 710986, "postDate": "2020-01-05T14:12:28.320Z", "rawMarkdown": "", "votes": 1, "isDeleted": true } ], "comments": [ { "id": 711005, "author_name": "Tim Yee", "author_url": "", "post_date": "2020-01-05T14:44:12.890000", "content": "This is a synchronous rerun code competition, you can assume that the complete test set will contain essentially the same size and number of images as the training set. Consider performing inference on just one batch at a time to avoid memory errors. Only the first few rows/images in the test set and sample submission files can be downloaded. These samples provided so you can review the basic structure of the files and to ensure consistency between the publicly available set of file names and those your code will have access to while it is being rerun for scoring.
\n
I don't think you can look at the sample submission to infer anything with respect to differences in LB scores since 12 test files is probably just enough to make sure your notebook runs correctly. It gets re-run on full test set once you hit submit.
\n\n\n", "votes": 3, "replies": [] }, { "id": 710986, "author_name": "", "author_url": "", "post_date": "2020-01-05T14:12:28.320000", "content": "", "votes": 1, "replies": [] } ], "raw_markdown_by_id": { "710008": "I took my 4 best models(`0.96+` on LB) and `submission.csv` from best public kernel(`0.9663+`) and compared the target values. This is what I got \n\n\nThey are pretty consistent excpet for one: `Test_6_grapheme_root`. \n\nDoes this mean that models with `0.97+` on LB will predict different values for `x` number of the row_id? What could be the possible `x`?\n", "711005": "I don't think you can look at the sample submission to infer anything with respect to differences in LB scores since 12 test files is probably just enough to make sure your notebook runs correctly. It gets re-run on full test set once you hit submit.\n\n> This is a synchronous rerun code competition, you can assume that the complete test set will contain essentially the same size and number of images as the training set. Consider performing inference on just one batch at a time to avoid memory errors. Only the first few rows/images in the test set and sample submission files can be downloaded. These samples provided so you can review the basic structure of the files and to ensure consistency between the publicly available set of file names and those your code will have access to while it is being rerun for scoring.", "710986": "" } }This is a synchronous rerun code competition, you can assume that the complete test set will contain essentially the same size and number of images as the training set. Consider performing inference on just one batch at a time to avoid memory errors. Only the first few rows/images in the test set and sample submission files can be downloaded. These samples provided so you can review the basic structure of the files and to ensure consistency between the publicly available set of file names and those your code will have access to while it is being rerun for scoring.
\n