It's official. Kernels can run for 6 hours (but only idle for 1 hour).
It seems that public LB scores for new and revised kernels are no longer appearing on the Kernels tab. (I posted in the Product Feedback forum already.) I don't know if this is a bug or meant to be a feature, but if it's an intended change, I wish they would get rid of the old scores too. Anyhow it messes up my Trump joke because sorting by score won't put my new blend kernel at the top. [EDIT: For a while it seemed that this was fixed, but now new scores are again no longer appearing on the Kernels tab.]
Oh, this is cool: we can now install custom packages in kernels.
It's official. Kernels can run for 6 hours (but only idle for 1 hour).
It seems that public LB scores for new and revised kernels are no longer appearing on the Kernels tab. (I posted in the Product Feedback forum already.) I don't know if this is a bug or meant to be a feature, but if it's an intended change, I wish they would get rid of the old scores too. Anyhow it messes up my Trump joke because sorting by score won't put my new blend kernel at the top. [EDIT: For a while it seemed that this was fixed, but now new scores are again no longer appearing on the Kernels tab.]
Oh, this is cool: we can now install custom packages in kernels.
Kernel seems getting better each day. I hope GPU or more cores and memory are next.
", "rawMarkdown": "Kernel seems getting better each day. I hope GPU or more cores and memory are next.", "votes": null }, { "id": "300963", "postDate": "03/22/2018 05:41:08", "content": "That's great news. I hope, memory limit will also be increased in coming future considering the size of this dataset.
\n\nRegarding point 2, I also noticed exactly the same bug. My kernel still shows old score. Situation is worst for R markdown report. I have just now created a kernel in R markdown which has prediction file + score but it doesn't show score in list of Kernels (Kernel's tab).
", "rawMarkdown": "That's great news. I hope, memory limit will also be increased in coming future considering the size of this dataset. \n\nRegarding point 2, I also noticed exactly the same bug. My [kernel][1] still shows old score. Situation is worst for R markdown report. I have just now created a kernel in R markdown which has prediction file + score but it doesn't show score in list of Kernels (Kernel's tab). \n\n\n [1]: https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection/kernels?sortBy=votes&group=everyone&pageSize=20&competitionId=8540", "votes": null }, { "id": "301132", "postDate": "03/22/2018 11:00:12", "content": "Perhaps actually point 2) is good for stopping the blind blending. Now you have to at least see what the script is more or less doing :)
", "rawMarkdown": "Perhaps actually point 2) is good for stopping the blind blending. Now you have to at least see what the script is more or less doing :)", "votes": null }, { "id": "305445", "postDate": "03/28/2018 22:10:43", "content": "Public LB scores are back on the Kernels tab. And right now I have the highest scoring single-model public kernel. (Actually I just translated Pranav Pandya's kernel from R to Python.) We'll see how long that lasts.
", "rawMarkdown": "Public LB scores are back on the Kernels tab. And right now I have the highest scoring [single-model public kernel][1]. (Actually I just translated [Pranav Pandya's kernel][2] from R to Python.) We'll see how long that lasts.\n\n\n [1]: https://www.kaggle.com/aharless/try-pranav-s-r-lgbm-in-python?scriptVersionId=2973530\n [2]: https://www.kaggle.com/pranav84/single-lightgbm-in-r-with-75-mln-rows-lb-0-9690", "votes": null }, { "id": "306372", "postDate": "03/30/2018 10:21:42", "content": "I meant \"we'll see how long my high score lasts,\" but it turns out that the fix for the Kernels tab has not lasted. New scores are no longer appearing.
", "rawMarkdown": "I meant \"we'll see how long my high score lasts,\" but it turns out that the fix for the Kernels tab has not lasted. New scores are no longer appearing.", "votes": null } ], "comments": [ { "id": 300922, "author_name": "muhammadalfiansyah", "author_url": "", "post_date": "03/22/2018 04:14:15", "content": "Kernel seems getting better each day. I hope GPU or more cores and memory are next.
", "votes": null, "replies": [] }, { "id": 300963, "author_name": "pranav84", "author_url": "", "post_date": "03/22/2018 05:41:08", "content": "That's great news. I hope, memory limit will also be increased in coming future considering the size of this dataset.
\n\nRegarding point 2, I also noticed exactly the same bug. My kernel still shows old score. Situation is worst for R markdown report. I have just now created a kernel in R markdown which has prediction file + score but it doesn't show score in list of Kernels (Kernel's tab).
", "votes": null, "replies": [] }, { "id": 301132, "author_name": "asparuhhristov", "author_url": "", "post_date": "03/22/2018 11:00:12", "content": "Perhaps actually point 2) is good for stopping the blind blending. Now you have to at least see what the script is more or less doing :)
", "votes": null, "replies": [] }, { "id": 305445, "author_name": "aharless", "author_url": "", "post_date": "03/28/2018 22:10:43", "content": "Public LB scores are back on the Kernels tab. And right now I have the highest scoring single-model public kernel. (Actually I just translated Pranav Pandya's kernel from R to Python.) We'll see how long that lasts.
", "votes": null, "replies": [ { "id": 306372, "author_name": "aharless", "author_url": "", "post_date": "03/30/2018 10:21:42", "content": "I meant \"we'll see how long my high score lasts,\" but it turns out that the fix for the Kernels tab has not lasted. New scores are no longer appearing.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "300860": "1. [It's official.][1] Kernels can run for 6 hours (but only idle for 1 hour).\n\n2. It seems that public LB scores for new and revised kernels are no longer appearing on the Kernels tab. (I posted in the [Product Feedback][2] forum already.)