![\"\"](\"https://i.imgflip.com/37ohix.jpg\")
Late at night, you're tired, and you don't create a new conda env when installing something. And you did not notice this install also installs a new Python version. From then, nothing works anymore, notebooks don't start, etc...
\n\nIt happened to me last night, and I spent several hours reinstalling a clean python stack.
\n\nSure, it could have been worse, it could have happened 2 days before competition end...
\n\nThe good side is that I have an up to date machine.
\n\nLet's go back to competing.
", "messages": [ { "id": 595455, "postDate": "2019-08-09T08:46:17.960Z", "content": "Late at night, you're tired, and you don't create a new conda env when installing something. And you did not notice this install also installs a new Python version. From then, nothing works anymore, notebooks don't start, etc...
\n\nIt happened to me last night, and I spent several hours reinstalling a clean python stack.
\n\nSure, it could have been worse, it could have happened 2 days before competition end...
\n\nThe good side is that I have an up to date machine.
\n\nLet's go back to competing.
", "rawMarkdown": "Late at night, you're tired, and you don't create a new conda env when installing something. And you did not notice this install also installs a new Python version. From then, nothing works anymore, notebooks don't start, etc...\n\nIt happened to me last night, and I spent several hours reinstalling a clean python stack. \n\nSure, it could have been worse, it could have happened 2 days before competition end...\n\nThe good side is that I have an up to date machine.\n\nLet's go back to competing.", "votes": 20 }, { "id": 596107, "postDate": "2019-08-10T07:15:31.467Z", "content": "
Well that is a very targeted reference 😁
", "rawMarkdown": "Well that is a very targeted reference 😁 ", "votes": 2 }, { "id": 596417, "postDate": "2019-08-10T16:03:41.103Z", "content": "I didn't dare install rdkit on my machine, I used Kaggle kernel when I used rdkit :)
", "rawMarkdown": "I didn't dare install rdkit on my machine, I used Kaggle kernel when I used rdkit :)" }, { "id": 596536, "postDate": "2019-08-10T19:50:47.297Z", "content": "Haha, I've had to clean install because of this comp. Pip + pip3 + conda2 + conda3 + rdkit + openbabel + bugs = need a new pc after.
", "rawMarkdown": "Haha, I've had to clean install because of this comp. Pip + pip3 + conda2 + conda3 + rdkit + openbabel + bugs = need a new pc after.", "votes": 4 }, { "id": 597439, "postDate": "2019-08-12T10:53:33.497Z", "content": "RDKIT is beautiful for visualization model of neural network Thank you for your share
", "rawMarkdown": "RDKIT is beautiful for visualization model of neural network Thank you for your share" } ] }, { "id": 595461, "postDate": "2019-08-09T08:55:33.250Z", "content": "Keep calm. \nBrew some coffee. \nWait for it's kicking in.
\n\n\n", "rawMarkdown": "Keep calm. \nBrew some coffee. \nWait for it's kicking in.\n\n> Let's go back to competing.", "votes": 3, "replies": [ { "id": 595556, "postDate": "2019-08-09T11:21:19.017Z", "content": "Let's go back to competing.
\n
I stayed calm, getting angry can just lead to even worse situations.
", "rawMarkdown": "I stayed calm, getting angry can just lead to even worse situations.", "votes": 3 } ] }, { "id": 596937, "postDate": "2019-08-11T14:42:25.623Z", "content": "Already done it! I ended up having to compile tensorflow from source an absolute joy - swearbox is now full ;(
\n\nI am not that impressed with this conda
", "rawMarkdown": "Already done it! I ended up having to compile tensorflow from source an absolute joy - swearbox is now full ;(\n\nI am not that impressed with this conda", "votes": 1, "replies": [ { "id": 597321, "postDate": "2019-08-12T07:13:13.027Z", "content": "Wow, for me it was simple: wipe anaconda, install latest, build environements with what I use. Everything with conda, except few pip install like for tensoflow-gpu. But I'm using ubuntu 18.04 which may make it simpler.
", "rawMarkdown": "Wow, for me it was simple: wipe anaconda, install latest, build environements with what I use. Everything with conda, except few pip install like for tensoflow-gpu. But I'm using ubuntu 18.04 which may make it simpler." } ] }, { "id": 595706, "postDate": "2019-08-09T15:06:23.343Z", "content": "Docker for life :)
", "rawMarkdown": "Docker for life :)", "votes": 1, "replies": [ { "id": 595758, "postDate": "2019-08-09T16:11:00.093Z", "content": "You create a new docker image each time you install or upgrade a package?
", "rawMarkdown": "You create a new docker image each time you install or upgrade a package?" }, { "id": 595759, "postDate": "2019-08-09T16:15:43.200Z", "content": "I have a Docker that I upgrade every 3 months. Then usually for each project I can do some light installation base on this image
", "rawMarkdown": "I have a Docker that I upgrade every 3 months. Then usually for each project I can do some light installation base on this image", "votes": 2 }, { "id": 595824, "postDate": "2019-08-09T18:28:09.037Z", "content": "mine was a rather light install: I upgraded numpy...
", "rawMarkdown": "mine was a rather light install: I upgraded numpy...", "votes": 1 } ] }, { "id": 595470, "postDate": "2019-08-09T09:05:53.937Z", "content": "If you use btrfs try snapper.io it has saved my life wiping my whole home directory
", "rawMarkdown": "If you use btrfs try snapper.io it has saved my life wiping my whole home directory", "votes": 1, "replies": [ { "id": 595559, "postDate": "2019-08-09T11:24:20.360Z", "content": "Thanks, I didn't know about btrfs nor about snapper.io. Will look into it.
", "rawMarkdown": "Thanks, I didn't know about btrfs nor about snapper.io. Will look into it." } ] }, { "id": 597872, "postDate": "2019-08-12T20:52:56.420Z", "content": "I can empathize:\nFor various reasons I prefer to use R as my \"base of operations\" for Kaggle contests rather than Python, although I can get to Python (3.6) from R if I need to via the \"reticulate\" package. In addition to other machine learning tools like lightgbm, xgboost, etc., R has packages that expose the keras and tensorflow APIs directly, although I think they use reticulate to accomplish this. It all works as long as the components (Ubuntu OS, my R installation, the R packages for keras and tensorflow, my Python installation, the Python modules for keras and tensorflow, and cuda for the GPU) all play nice together. But an innocent looking but incompatible update to any one of these can break the toolchain. I ran into this several days ago with keras and tensorflow, which started issuing mysterious error messages about class and method incompatibilities involving my model objects. After many hours of testing, updating, and more testing, including installing a development Github version of R keras, I finally got the whole thing working again. During that pitched and gory battle there were dark times when I wasn't sure I would emerge the victor, but somewhere inside of me I knew from long experience that persistence and grit should eventually win out. Reminds me of that expression: \"What doesn't kill you makes you stronger\".
", "rawMarkdown": "I can empathize:\nFor various reasons I prefer to use R as my \"base of operations\" for Kaggle contests rather than Python, although I can get to Python (3.6) from R if I need to via the \"reticulate\" package. In addition to other machine learning tools like lightgbm, xgboost, etc., R has packages that expose the keras and tensorflow APIs directly, although I think they use reticulate to accomplish this. It all works as long as the components (Ubuntu OS, my R installation, the R packages for keras and tensorflow, my Python installation, the Python modules for keras and tensorflow, and cuda for the GPU) all play nice together. But an innocent looking but incompatible update to any one of these can break the toolchain. I ran into this several days ago with keras and tensorflow, which started issuing mysterious error messages about class and method incompatibilities involving my model objects. After many hours of testing, updating, and more testing, including installing a development Github version of R keras, I finally got the whole thing working again. During that pitched and gory battle there were dark times when I wasn't sure I would emerge the victor, but somewhere inside of me I knew from long experience that persistence and grit should eventually win out. Reminds me of that expression: \"What doesn't kill you makes you stronger\".", "votes": 2 }, { "id": 595986, "postDate": "2019-08-10T00:18:09.620Z", "content": "I forced myself to use kaggle dockers and pull their latest changes. Of course we have an internal version with our rapids software. Here is our PR https://github.com/Kaggle/docker-python/pull/595
", "rawMarkdown": "I forced myself to use kaggle dockers and pull their latest changes. Of course we have an internal version with our rapids software. Here is our PR https://github.com/Kaggle/docker-python/pull/595", "votes": 2 }, { "id": 595510, "postDate": "2019-08-09T09:55:17.830Z", "content": "Sorry to hear that @cpmpml ... but a fresh install may mean more upside subs to come for you guys ;-)
", "rawMarkdown": "Sorry to hear that @cpmpml ... but a fresh install may mean more upside subs to come for you guys ;-) ", "replies": [ { "id": 595560, "postDate": "2019-08-09T11:25:21.137Z", "content": "Thanks. If only refreshing versions were enough to improve model performances... ;)
", "rawMarkdown": "Thanks. If only refreshing versions were enough to improve model performances... ;)", "votes": 1 } ] }, { "id": 597674, "postDate": "2019-08-12T16:40:40.063Z", "content": "Sorry for being off topic, but what IDE do you people use for Python?
", "rawMarkdown": "Sorry for being off topic, but what IDE do you people use for Python?", "replies": [ { "id": 597684, "postDate": "2019-08-12T16:54:29.973Z", "content": "Currently, I use just Jupyter notebooks and Vim.
", "rawMarkdown": "Currently, I use just Jupyter notebooks and Vim.", "votes": 1 } ] }, { "id": 596496, "postDate": "2019-08-10T18:13:01.157Z", "content": "I haven't done this yet, but it's definitely possible.
", "rawMarkdown": "I haven't done this yet, but it's definitely possible." }, { "id": 595946, "postDate": "2019-08-09T22:22:24.530Z", "content": "happens!
", "rawMarkdown": "happens!" }, { "id": 598032, "postDate": "2019-08-13T03:09:16.480Z", "rawMarkdown": "", "isDeleted": true }, { "id": 595744, "postDate": "2019-08-09T15:49:31.683Z", "rawMarkdown": "", "votes": 4, "isDeleted": true, "replies": [ { "id": 595757, "postDate": "2019-08-09T16:10:27.290Z", "content": "Exactly, I have no excuse ;)
", "rawMarkdown": "Exactly, I have no excuse ;)" }, { "id": 595972, "postDate": "2019-08-09T23:16:47.733Z", "content": "The irony is that a similar effect you can obtain updating conda. The only difference is that someone who does not like Python does not repair it in several hours.
", "rawMarkdown": "The irony is that a similar effect you can obtain updating conda. The only difference is that someone who does not like Python does not repair it in several hours.", "votes": 2 } ] } ], "comments": [ { "id": 596107, "author_name": "Dieter", "author_url": "", "post_date": "2019-08-10T07:15:31.467000", "content": "
Well that is a very targeted reference 😁
", "votes": 2, "replies": [] }, { "id": 596417, "author_name": "CPMP", "author_url": "", "post_date": "2019-08-10T16:03:41.103000", "content": "I didn't dare install rdkit on my machine, I used Kaggle kernel when I used rdkit :)
", "votes": 0, "replies": [] }, { "id": 596536, "author_name": "roush", "author_url": "", "post_date": "2019-08-10T19:50:47.297000", "content": "Haha, I've had to clean install because of this comp. Pip + pip3 + conda2 + conda3 + rdkit + openbabel + bugs = need a new pc after.
", "votes": 4, "replies": [] }, { "id": 597439, "author_name": "ForesightKing", "author_url": "", "post_date": "2019-08-12T10:53:33.497000", "content": "RDKIT is beautiful for visualization model of neural network Thank you for your share
", "votes": 0, "replies": [] } ] }, { "id": 595461, "author_name": "Stanislav Blinov", "author_url": "", "post_date": "2019-08-09T08:55:33.250000", "content": "Keep calm. \nBrew some coffee. \nWait for it's kicking in.
\n\n\n", "votes": 3, "replies": [ { "id": 595556, "author_name": "CPMP", "author_url": "", "post_date": "2019-08-09T11:21:19.017000", "content": "Let's go back to competing.
\n
I stayed calm, getting angry can just lead to even worse situations.
", "votes": 3, "replies": [] } ] }, { "id": 596937, "author_name": "Scirpus", "author_url": "", "post_date": "2019-08-11T14:42:25.623000", "content": "Already done it! I ended up having to compile tensorflow from source an absolute joy - swearbox is now full ;(
\n\nI am not that impressed with this conda
", "votes": 1, "replies": [ { "id": 597321, "author_name": "CPMP", "author_url": "", "post_date": "2019-08-12T07:13:13.027000", "content": "Wow, for me it was simple: wipe anaconda, install latest, build environements with what I use. Everything with conda, except few pip install like for tensoflow-gpu. But I'm using ubuntu 18.04 which may make it simpler.
", "votes": 0, "replies": [] } ] }, { "id": 595706, "author_name": "Kaggoo", "author_url": "", "post_date": "2019-08-09T15:06:23.343000", "content": "Docker for life :)
", "votes": 1, "replies": [ { "id": 595758, "author_name": "CPMP", "author_url": "", "post_date": "2019-08-09T16:11:00.093000", "content": "You create a new docker image each time you install or upgrade a package?
", "votes": 0, "replies": [] }, { "id": 595759, "author_name": "Lam Dang", "author_url": "", "post_date": "2019-08-09T16:15:43.200000", "content": "I have a Docker that I upgrade every 3 months. Then usually for each project I can do some light installation base on this image
", "votes": 2, "replies": [] }, { "id": 595824, "author_name": "CPMP", "author_url": "", "post_date": "2019-08-09T18:28:09.037000", "content": "mine was a rather light install: I upgraded numpy...
", "votes": 1, "replies": [] } ] }, { "id": 595470, "author_name": "Andrés Miguel Torrubia Sáez", "author_url": "", "post_date": "2019-08-09T09:05:53.937000", "content": "If you use btrfs try snapper.io it has saved my life wiping my whole home directory
", "votes": 1, "replies": [ { "id": 595559, "author_name": "CPMP", "author_url": "", "post_date": "2019-08-09T11:24:20.360000", "content": "Thanks, I didn't know about btrfs nor about snapper.io. Will look into it.
", "votes": 0, "replies": [] } ] }, { "id": 597872, "author_name": "David J. Slate", "author_url": "", "post_date": "2019-08-12T20:52:56.420000", "content": "I can empathize:\nFor various reasons I prefer to use R as my \"base of operations\" for Kaggle contests rather than Python, although I can get to Python (3.6) from R if I need to via the \"reticulate\" package. In addition to other machine learning tools like lightgbm, xgboost, etc., R has packages that expose the keras and tensorflow APIs directly, although I think they use reticulate to accomplish this. It all works as long as the components (Ubuntu OS, my R installation, the R packages for keras and tensorflow, my Python installation, the Python modules for keras and tensorflow, and cuda for the GPU) all play nice together. But an innocent looking but incompatible update to any one of these can break the toolchain. I ran into this several days ago with keras and tensorflow, which started issuing mysterious error messages about class and method incompatibilities involving my model objects. After many hours of testing, updating, and more testing, including installing a development Github version of R keras, I finally got the whole thing working again. During that pitched and gory battle there were dark times when I wasn't sure I would emerge the victor, but somewhere inside of me I knew from long experience that persistence and grit should eventually win out. Reminds me of that expression: \"What doesn't kill you makes you stronger\".
", "votes": 2, "replies": [] }, { "id": 595986, "author_name": "Jiwei Liu", "author_url": "", "post_date": "2019-08-10T00:18:09.620000", "content": "I forced myself to use kaggle dockers and pull their latest changes. Of course we have an internal version with our rapids software. Here is our PR https://github.com/Kaggle/docker-python/pull/595
", "votes": 2, "replies": [] }, { "id": 595510, "author_name": "olivier", "author_url": "", "post_date": "2019-08-09T09:55:17.830000", "content": "Sorry to hear that @cpmpml ... but a fresh install may mean more upside subs to come for you guys ;-)
", "votes": 0, "replies": [ { "id": 595560, "author_name": "CPMP", "author_url": "", "post_date": "2019-08-09T11:25:21.137000", "content": "Thanks. If only refreshing versions were enough to improve model performances... ;)
", "votes": 1, "replies": [] } ] }, { "id": 597674, "author_name": "agenlu", "author_url": "", "post_date": "2019-08-12T16:40:40.063000", "content": "Sorry for being off topic, but what IDE do you people use for Python?
", "votes": 0, "replies": [ { "id": 597684, "author_name": "somnambWl", "author_url": "", "post_date": "2019-08-12T16:54:29.973000", "content": "Currently, I use just Jupyter notebooks and Vim.
", "votes": 1, "replies": [] } ] }, { "id": 596496, "author_name": "Ingus T", "author_url": "", "post_date": "2019-08-10T18:13:01.157000", "content": "I haven't done this yet, but it's definitely possible.
", "votes": 0, "replies": [] }, { "id": 595946, "author_name": "Firat Gonen", "author_url": "", "post_date": "2019-08-09T22:22:24.530000", "content": "happens!
", "votes": 0, "replies": [] }, { "id": 598032, "author_name": "", "author_url": "", "post_date": "2019-08-13T03:09:16.480000", "content": "", "votes": 0, "replies": [] }, { "id": 595744, "author_name": "", "author_url": "", "post_date": "2019-08-09T15:49:31.683000", "content": "", "votes": 4, "replies": [ { "id": 595757, "author_name": "CPMP", "author_url": "", "post_date": "2019-08-09T16:10:27.290000", "content": "Exactly, I have no excuse ;)
", "votes": 0, "replies": [] }, { "id": 595972, "author_name": "Grzegorz Sionkowski", "author_url": "", "post_date": "2019-08-09T23:16:47.733000", "content": "The irony is that a similar effect you can obtain updating conda. The only difference is that someone who does not like Python does not repair it in several hours.
", "votes": 2, "replies": [] } ] } ], "raw_markdown_by_id": { "595455": "Late at night, you're tired, and you don't create a new conda env when installing something. And you did not notice this install also installs a new Python version. From then, nothing works anymore, notebooks don't start, etc...\n\nIt happened to me last night, and I spent several hours reinstalling a clean python stack. \n\nSure, it could have been worse, it could have happened 2 days before competition end...\n\nThe good side is that I have an up to date machine.\n\nLet's go back to competing.", "596107": "", "595461": "Keep calm. \nBrew some coffee. \nWait for it's kicking in.\n\n> Let's go back to competing.", "596937": "Already done it! I ended up having to compile tensorflow from source an absolute joy - swearbox is now full ;(\n\nI am not that impressed with this conda", "595706": "Docker for life :)", "595470": "If you use btrfs try snapper.io it has saved my life wiping my whole home directory", "597872": "I can empathize:\nFor various reasons I prefer to use R as my \"base of operations\" for Kaggle contests rather than Python, although I can get to Python (3.6) from R if I need to via the \"reticulate\" package. In addition to other machine learning tools like lightgbm, xgboost, etc., R has packages that expose the keras and tensorflow APIs directly, although I think they use reticulate to accomplish this. It all works as long as the components (Ubuntu OS, my R installation, the R packages for keras and tensorflow, my Python installation, the Python modules for keras and tensorflow, and cuda for the GPU) all play nice together. But an innocent looking but incompatible update to any one of these can break the toolchain. I ran into this several days ago with keras and tensorflow, which started issuing mysterious error messages about class and method incompatibilities involving my model objects. After many hours of testing, updating, and more testing, including installing a development Github version of R keras, I finally got the whole thing working again. During that pitched and gory battle there were dark times when I wasn't sure I would emerge the victor, but somewhere inside of me I knew from long experience that persistence and grit should eventually win out. Reminds me of that expression: \"What doesn't kill you makes you stronger\".", "595986": "I forced myself to use kaggle dockers and pull their latest changes. Of course we have an internal version with our rapids software. Here is our PR https://github.com/Kaggle/docker-python/pull/595", "595510": "Sorry to hear that @cpmpml ... but a fresh install may mean more upside subs to come for you guys ;-) ", "597674": "Sorry for being off topic, but what IDE do you people use for Python?", "596496": "I haven't done this yet, but it's definitely possible.", "595946": "happens!", "598032": "", "595744": "" } }