So, in my team case, our submissions inferences are way too fast for some reason like 5 minutes to calculate the score, However it takes 15-30 minutes (depending on the model) to commit the inference and calculate the values for 400 videos, while the test set is 4000 videos, that is weird, anyone encountered this problem yet?
\n\nMaybe Kaggle Team should answer...
", "messages": [ { "id": 729258, "postDate": "2020-01-26T00:32:05.943Z", "content": "So, in my team case, our submissions inferences are way too fast for some reason like 5 minutes to calculate the score, However it takes 15-30 minutes (depending on the model) to commit the inference and calculate the values for 400 videos, while the test set is 4000 videos, that is weird, anyone encountered this problem yet?
\n\nMaybe Kaggle Team should answer...
", "rawMarkdown": "So, in my team case, our submissions inferences are way too fast for some reason like 5 minutes to calculate the score, However it takes 15-30 minutes (depending on the model) to commit the inference and calculate the values for 400 videos, while the test set is 4000 videos, that is weird, anyone encountered this problem yet?\n\nMaybe Kaggle Team should answer...", "votes": 7 }, { "id": 729363, "postDate": "2020-01-26T05:02:02.057Z", "content": "Even if one code block in jupyter notebook fails, next block is executed. Maybe your inference code block fails after some videos. Try try except when reading video since there can be corrupt videos in test set.
FYI We already solved this so...
", "rawMarkdown": "FYI We already solved this so..." }, { "id": 738942, "postDate": "2020-02-07T07:42:48.993Z", "content": "it was calculated by probing. You can see details in one of the discussion threads. However, it is unwise to use numbers as the final scoring will be on a completely different set.
", "rawMarkdown": "it was calculated by probing. You can see details in one of the discussion threads. However, it is unwise to use numbers as the final scoring will be on a completely different set." }, { "id": 738826, "postDate": "2020-02-07T04:17:58.373Z", "content": "Thanks for sharing！@Harshit Sheoran Can you provide some details about the “try except” . I want to learn and discuss the specific operation methods in \"except\".
", "rawMarkdown": "Thanks for sharing！@Harshit Sheoran Can you provide some details about the “try except” . I want to learn and discuss the specific operation methods in \"except\".", "replies": [ { "id": 738938, "postDate": "2020-02-07T07:32:06.663Z", "content": "So, try except is a block, it works like, if the code under try receives an error, it will not raise the error and stop the program from running, but instead, it will go ahead and execute the code under except block, if the code in the try runs without and error, the except block is not executed, however if the code under try and except both blocks raises an error, the code will rasie error while executing except block and stop.
", "rawMarkdown": "So, try except is a block, it works like, if the code under try receives an error, it will not raise the error and stop the program from running, but instead, it will go ahead and execute the code under except block, if the code in the try runs without and error, the except block is not executed, however if the code under try and except both blocks raises an error, the code will rasie error while executing except block and stop." }, { "id": 739066, "postDate": "2020-02-07T11:18:54.647Z", "content": "Thank you for reply
", "rawMarkdown": "Thank you for reply" }, { "id": 739432, "postDate": "2020-02-07T20:45:00.587Z", "content": "\ntry:\n vid = [load video]\n img = [detect face]\n pred = [predict on model]\nexcept:\n pred = 0.5\n\nPublic kernal that implements it: https://www.kaggle.com/humananalog/inference-demo
It takes 35 minutes to run on 400 videos while commiting, then lb score is calculated in 5 minutes after submitting the submission file... How?
", "rawMarkdown": "It takes 35 minutes to run on 400 videos while commiting, then lb score is calculated in 5 minutes after submitting the submission file... How?", "votes": 1 }, { "id": 729401, "postDate": "2020-01-26T06:22:36.540Z", "content": "I misunderstood your question. \nThe public test set is not exactly same as public validation set (frame count, resolution...). which will cause failure.
", "rawMarkdown": "I misunderstood your question. \nThe public test set is not exactly same as public validation set (frame count, resolution...). which will cause failure." }, { "id": 729710, "postDate": "2020-01-26T14:49:40.987Z", "content": "Yeah that might be the problem, but with all of my inferences, I mean I have different inferences, some of them are forked from kaggle kernels and I just put my model in it, But it still does not work.
", "rawMarkdown": "Yeah that might be the problem, but with all of my inferences, I mean I have different inferences, some of them are forked from kaggle kernels and I just put my model in it, But it still does not work." } ] } ], "comments": [ { "id": 729363, "author_name": "YoonSoo", "author_url": "", "post_date": "2020-01-26T05:02:02.057000", "content": "Even if one code block in jupyter notebook fails, next block is executed. Maybe your inference code block fails after some videos. Try try except when reading video since there can be corrupt videos in test set.
FYI We already solved this so...
", "votes": 0, "replies": [] }, { "id": 738942, "author_name": "Moshel", "author_url": "", "post_date": "2020-02-07T07:42:48.993000", "content": "it was calculated by probing. You can see details in one of the discussion threads. However, it is unwise to use numbers as the final scoring will be on a completely different set.
", "votes": 0, "replies": [] }, { "id": 738826, "author_name": "beiweizai", "author_url": "", "post_date": "2020-02-07T04:17:58.373000", "content": "Thanks for sharing！@Harshit Sheoran Can you provide some details about the “try except” . I want to learn and discuss the specific operation methods in \"except\".
", "votes": 0, "replies": [ { "id": 738938, "author_name": "Harshit Sheoran", "author_url": "", "post_date": "2020-02-07T07:32:06.663000", "content": "So, try except is a block, it works like, if the code under try receives an error, it will not raise the error and stop the program from running, but instead, it will go ahead and execute the code under except block, if the code in the try runs without and error, the except block is not executed, however if the code under try and except both blocks raises an error, the code will rasie error while executing except block and stop.
", "votes": 0, "replies": [] }, { "id": 739066, "author_name": "beiweizai", "author_url": "", "post_date": "2020-02-07T11:18:54.647000", "content": "Thank you for reply
", "votes": 0, "replies": [] }, { "id": 739432, "author_name": "GreatGameDota", "author_url": "", "post_date": "2020-02-07T20:45:00.587000", "content": "\ntry:\n vid = [load video]\n img = [detect face]\n pred = [predict on model]\nexcept:\n pred = 0.5\n\nPublic kernal that implements it: https://www.kaggle.com/humananalog/inference-demo
It takes 35 minutes to run on 400 videos while commiting, then lb score is calculated in 5 minutes after submitting the submission file... How?
", "votes": 1, "replies": [] }, { "id": 729401, "author_name": "Johnny Lee", "author_url": "", "post_date": "2020-01-26T06:22:36.540000", "content": "I misunderstood your question. \nThe public test set is not exactly same as public validation set (frame count, resolution...). which will cause failure.
", "votes": 0, "replies": [] }, { "id": 729710, "author_name": "Harshit Sheoran", "author_url": "", "post_date": "2020-01-26T14:49:40.987000", "content": "Yeah that might be the problem, but with all of my inferences, I mean I have different inferences, some of them are forked from kaggle kernels and I just put my model in it, But it still does not work.
", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "729258": "So, in my team case, our submissions inferences are way too fast for some reason like 5 minutes to calculate the score, However it takes 15-30 minutes (depending on the model) to commit the inference and calculate the values for 400 videos, while the test set is 4000 videos, that is weird, anyone encountered this problem yet?\n\nMaybe Kaggle Team should answer...", "729363": "Even if one code block in jupyter notebook fails, next block is executed. Maybe your inference code block fails after some videos. Try `try except` when reading video since there can be corrupt videos in test set.", "739551": "FYI We already solved this so...", "738942": "it was calculated by probing. You can see details in one of the discussion threads. However, it is unwise to use numbers as the final scoring will be on a completely different set.", "738826": "Thanks for sharing！@Harshit Sheoran Can you provide some details about the “try except” . I want to learn and discuss the specific operation methods in \"except\".", "729316": "" } }