A lot of kernels are not going to run in the second stage due to time constraints. \nOne thing that you can do to simulate if your kernel is going to fail is to replicate your test data 7(376k/56k) times using something like :
\n\ntest_df = pd.concat([test_data]*7)\nAnd running your kernel. \nIf you are able to submit to competition, you are most probably good to go.
", "messages": [ { "id": "446741", "postDate": "12/28/2018 16:00:46", "content": "A lot of kernels are not going to run in the second stage due to time constraints. \nOne thing that you can do to simulate if your kernel is going to fail is to replicate your test data 7(376k/56k) times using something like :
\n\ntest_df = pd.concat([test_data]*7)\nAnd running your kernel. \nIf you are able to submit to competition, you are most probably good to go.
", "rawMarkdown": "A lot of kernels are not going to run in the second stage due to time constraints. \nOne thing that you can do to simulate if your kernel is going to fail is to replicate your test data 7(376k/56k) times using something like :\n\n test_df = pd.concat([test_data]*7)\n\nAnd running your kernel. \nIf you are able to submit to competition, you are most probably good to go.", "votes": null }, { "id": "447063", "postDate": "12/29/2018 04:42:10", "content": "Sometimes my kernel runs about 7300s but the result can be submited. I doubt that the 2 hours limit is not absolute.
", "rawMarkdown": "Sometimes my kernel runs about 7300s but the result can be submited. I doubt that the 2 hours limit is not absolute.", "votes": null }, { "id": "447126", "postDate": "12/29/2018 07:29:25", "content": "Same for me. I have seen that if its less than 124 mins it will still be allowed to submit.
", "rawMarkdown": "Same for me. I have seen that if its less than 124 mins it will still be allowed to submit.", "votes": null }, { "id": "447241", "postDate": "12/29/2018 12:41:26", "content": "I think organizers will increase the time limit.
\n\nUpd:
\n\nKernels will need to take into consideration the additional processing and inference time of the larger Test dataset. That will need to fit into the 2 / 6 hour constraints.\nhttps://www.kaggle.com/c/quora-insincere-questions-classification/discussion/70715#441672
", "rawMarkdown": "I think organizers will increase the time limit.\n\nUpd:\n\n Kernels will need to take into consideration the additional processing and inference time of the larger Test dataset. That will need to fit into the 2 / 6 hour constraints.\n\nhttps://www.kaggle.com/c/quora-insincere-questions-classification/discussion/70715#441672", "votes": null }, { "id": "447294", "postDate": "12/29/2018 15:27:45", "content": "Well, that statement clearly says they won't.
", "rawMarkdown": "Well, that statement clearly says they won't.", "votes": null }, { "id": "447577", "postDate": "12/30/2018 05:23:18", "content": "I think you need to calculate 7 * (preprocess time for test data) in order to figure out the exact time you need to execute the script.
", "rawMarkdown": "I think you need to calculate 7 * (preprocess time for test data) in order to figure out the exact time you need to execute the script.", "votes": null }, { "id": "447703", "postDate": "12/30/2018 11:20:20", "content": "The main problem with using this approach is that the time is not only taken in preprocessing. As you add more test data your models might also take more time to predict.
", "rawMarkdown": "The main problem with using this approach is that the time is not only taken in preprocessing. As you add more test data your models might also take more time to predict.", "votes": null } ], "comments": [ { "id": 447063, "author_name": "kaggleczs", "author_url": "", "post_date": "12/29/2018 04:42:10", "content": "Sometimes my kernel runs about 7300s but the result can be submited. I doubt that the 2 hours limit is not absolute.
", "votes": null, "replies": [ { "id": 447126, "author_name": "mlwhiz", "author_url": "", "post_date": "12/29/2018 07:29:25", "content": "Same for me. I have seen that if its less than 124 mins it will still be allowed to submit.
", "votes": null, "replies": [] } ] }, { "id": 447241, "author_name": "", "author_url": "", "post_date": "12/29/2018 12:41:26", "content": "I think organizers will increase the time limit.
\n\nUpd:
\n\nKernels will need to take into consideration the additional processing and inference time of the larger Test dataset. That will need to fit into the 2 / 6 hour constraints.\nhttps://www.kaggle.com/c/quora-insincere-questions-classification/discussion/70715#441672
", "votes": null, "replies": [ { "id": 447294, "author_name": "philippsinger", "author_url": "", "post_date": "12/29/2018 15:27:45", "content": "Well, that statement clearly says they won't.
", "votes": null, "replies": [] } ] }, { "id": 447577, "author_name": "hidehisaarai1213", "author_url": "", "post_date": "12/30/2018 05:23:18", "content": "I think you need to calculate 7 * (preprocess time for test data) in order to figure out the exact time you need to execute the script.
", "votes": null, "replies": [ { "id": 447703, "author_name": "mlwhiz", "author_url": "", "post_date": "12/30/2018 11:20:20", "content": "The main problem with using this approach is that the time is not only taken in preprocessing. As you add more test data your models might also take more time to predict.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "446741": "A lot of kernels are not going to run in the second stage due to time constraints. \nOne thing that you can do to simulate if your kernel is going to fail is to replicate your test data 7(376k/56k) times using something like :\n\n test_df = pd.concat([test_data]*7)\n\nAnd running your kernel. \nIf you are able to submit to competition, you are most probably good to go.", "447063": "Sometimes my kernel runs about 7300s but the result can be submited. I doubt that the 2 hours limit is not absolute.", "447126": "Same for me. I have seen that if its less than 124 mins it will still be allowed to submit.", "447241": "I think organizers will increase the time limit.\n\nUpd:\n\n Kernels will need to take into consideration the additional processing and inference time of the larger Test dataset. That will need to fit into the 2 / 6 hour constraints.\n\nhttps://www.kaggle.com/c/quora-insincere-questions-classification/discussion/70715#441672", "447294": "Well, that statement clearly says they won't.", "447577": "I think you need to calculate 7 * (preprocess time for test data) in order to figure out the exact time you need to execute the script.", "447703": "The main problem with using this approach is that the time is not only taken in preprocessing. As you add more test data your models might also take more time to predict." }, "source": "meta" }