\nhttps://www.kaggle.com/code/markwijkhuizen/aslfr-transformer-training-inference/log
\nyou can see the \"Environment\" on the log page, which is the image of this match.
\n
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3777797%2F3abceea34235d56ccc126ece80bc9e8e%2F1688970457864.png?generation=1688970578354774&alt=media\")
Hey Everyone,
\nI wanted to work on my local machine, yet I keep having issues with package conflicts :( I was wondering if anyone else had a similar experience, and if so share what you did to resolve?
", "messages": [ { "id": 2334202, "postDate": "2023-07-07T14:08:09.967Z", "content": "Hey Everyone,
\nI wanted to work on my local machine, yet I keep having issues with package conflicts :( I was wondering if anyone else had a similar experience, and if so share what you did to resolve?
", "rawMarkdown": "Hey Everyone,\n\nI wanted to work on my local machine, yet I keep having issues with package conflicts :( I was wondering if anyone else had a similar experience, and if so share what you did to resolve?", "votes": 1 }, { "id": 2334624, "postDate": "2023-07-07T20:32:16.487Z", "content": "Configuring the environment is quite troublesome. So I prefer docker, which means using someone else's environment. Here is the google image used for this match:
\nhttps://console.cloud.google.com/gcr/images/kaggle-gpu-images/GLOBAL/python@sha256:8a151f075ccff2e272bcada1d6ae32f4f527fdd1aadcb710f25c2eb557e91d55/details?tab=info
\nSupposed you have installed docker, then you can pull the image by:
\ndocker pull gcr.io/kaggle-gpu-images/python:v133
\nBut it is very big, taking 50.2GB.
I agree @chellyfan, using a docker image like this is a better alternative than trying to configure the environment on one's own.
", "rawMarkdown": "I agree @chellyfan, using a docker image like this is a better alternative than trying to configure the environment on one's own. ", "votes": 2 }, { "id": 2337483, "postDate": "2023-07-10T06:14:23.467Z", "content": "Thanks a bunch mate! I actually struggled and managed to get it going a little after my initial post, but if it throws me trouble later on, I'll just switch to the image!
\nOn a side note, how did you find it? I tried looking for an environment file / image myself but couldn't find it.
", "rawMarkdown": "Thanks a bunch mate! I actually struggled and managed to get it going a little after my initial post, but if it throws me trouble later on, I'll just switch to the image!\n\nOn a side note, how did you find it? I tried looking for an environment file / image myself but couldn't find it.", "replies": [ { "id": 2337500, "postDate": "2023-07-10T06:30:02.257Z", "content": "Actually, I found it by chance when I was browsing notebooks posted by others. For example, in this publicly available notebook:
\nhttps://www.kaggle.com/code/markwijkhuizen/aslfr-transformer-training-inference/log
\nyou can see the \"Environment\" on the log page, which is the image of this match.
\n
Thank you this is super useful!
", "rawMarkdown": "Thank you this is super useful!" }, { "id": 2339763, "postDate": "2023-07-11T04:35:00.183Z", "content": "Indeed thanks a bunch mate!
", "rawMarkdown": "Indeed thanks a bunch mate!" } ] } ] }, { "id": 2369551, "postDate": "2023-08-01T20:02:15.043Z", "content": "I just posted a new question, and maybe should have replied here first, so I'll do both. My new question: https://www.kaggle.com/competitions/asl-fingerspelling/discussion/428543
\nbut essentially it's that I tried to do what's here, but I'm using a Mac with M1 chip and I don't think it will work because it's not an Nvidia GPU? If I'm wrong, and it's possible, and anyone wants to help me figure out how to do it correctly, I'm all ears. Thank you
Hey everyone!
\nFirslty thanks @Zane for showing how to get the docker images that we can use. I also managed to get a functional environment going (or at least functional enough to run the introductory notebook). The trick is to use python venv instead of conda as there are a few packages that aren't available in conda's channels (most importantly mediapipe), so you would have to use pip to install them anyway. I am not quite sure why this is a problem, but it seems to cause issues with resolving versioning, leading to a lot of seeming conflicts.
\nI attached my requirements.txt file here if anyone wants to recreate for themselves, not that since I am running on Apple silicon, I also installed tensorflow-metal (and tf-macos but that was installed on it's own by a separate wheel).
\nHope this helps anyone else trying to run on their personal machine!
", "rawMarkdown": "Hey everyone! \n\nFirslty thanks @Zane for showing how to get the docker images that we can use. I also managed to get a functional environment going (or at least functional enough to run the introductory notebook). The trick is to use python venv instead of conda as there are a few packages that aren't available in conda's channels (most importantly mediapipe), so you would have to use pip to install them anyway. I am not quite sure why this is a problem, but it seems to cause issues with resolving versioning, leading to a lot of seeming conflicts. \n\nI attached my requirements.txt file here if anyone wants to recreate for themselves, not that since I am running on Apple silicon, I also installed tensorflow-metal (and tf-macos but that was installed on it's own by a separate wheel). \n\nHope this helps anyone else trying to run on their personal machine!" }, { "id": 2334527, "postDate": "2023-07-07T18:37:05.477Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 2334624, "author_name": "Zane", "author_url": "", "post_date": "2023-07-07T20:32:16.487000", "content": "Configuring the environment is quite troublesome. So I prefer docker, which means using someone else's environment. Here is the google image used for this match:
\nhttps://console.cloud.google.com/gcr/images/kaggle-gpu-images/GLOBAL/python@sha256:8a151f075ccff2e272bcada1d6ae32f4f527fdd1aadcb710f25c2eb557e91d55/details?tab=info
\nSupposed you have installed docker, then you can pull the image by:
\ndocker pull gcr.io/kaggle-gpu-images/python:v133
\nBut it is very big, taking 50.2GB.
I agree @chellyfan, using a docker image like this is a better alternative than trying to configure the environment on one's own.
", "votes": 2, "replies": [] }, { "id": 2337483, "author_name": "Vishesh Sarin", "author_url": "", "post_date": "2023-07-10T06:14:23.467000", "content": "Thanks a bunch mate! I actually struggled and managed to get it going a little after my initial post, but if it throws me trouble later on, I'll just switch to the image!
\nOn a side note, how did you find it? I tried looking for an environment file / image myself but couldn't find it.
", "votes": 0, "replies": [ { "id": 2337500, "author_name": "Zane", "author_url": "", "post_date": "2023-07-10T06:30:02.257000", "content": "Actually, I found it by chance when I was browsing notebooks posted by others. For example, in this publicly available notebook:
\nhttps://www.kaggle.com/code/markwijkhuizen/aslfr-transformer-training-inference/log
\nyou can see the \"Environment\" on the log page, which is the image of this match.
\n
Thank you this is super useful!
", "votes": 0, "replies": [] }, { "id": 2339763, "author_name": "Vishesh Sarin", "author_url": "", "post_date": "2023-07-11T04:35:00.183000", "content": "Indeed thanks a bunch mate!
", "votes": 0, "replies": [] } ] } ] }, { "id": 2369551, "author_name": "Kyle Proffitt", "author_url": "", "post_date": "2023-08-01T20:02:15.043000", "content": "I just posted a new question, and maybe should have replied here first, so I'll do both. My new question: https://www.kaggle.com/competitions/asl-fingerspelling/discussion/428543
\nbut essentially it's that I tried to do what's here, but I'm using a Mac with M1 chip and I don't think it will work because it's not an Nvidia GPU? If I'm wrong, and it's possible, and anyone wants to help me figure out how to do it correctly, I'm all ears. Thank you
Hey everyone!
\nFirslty thanks @Zane for showing how to get the docker images that we can use. I also managed to get a functional environment going (or at least functional enough to run the introductory notebook). The trick is to use python venv instead of conda as there are a few packages that aren't available in conda's channels (most importantly mediapipe), so you would have to use pip to install them anyway. I am not quite sure why this is a problem, but it seems to cause issues with resolving versioning, leading to a lot of seeming conflicts.
\nI attached my requirements.txt file here if anyone wants to recreate for themselves, not that since I am running on Apple silicon, I also installed tensorflow-metal (and tf-macos but that was installed on it's own by a separate wheel).
\nHope this helps anyone else trying to run on their personal machine!
", "votes": 0, "replies": [] }, { "id": 2334527, "author_name": "", "author_url": "", "post_date": "2023-07-07T18:37:05.477000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "2334202": "Hey Everyone,\n\nI wanted to work on my local machine, yet I keep having issues with package conflicts :( I was wondering if anyone else had a similar experience, and if so share what you did to resolve?", "2334624": "Configuring the environment is quite troublesome. So I prefer docker, which means using someone else's environment. Here is the google image used for this match:\nhttps://console.cloud.google.com/gcr/images/kaggle-gpu-images/GLOBAL/python@sha256:8a151f075ccff2e272bcada1d6ae32f4f527fdd1aadcb710f25c2eb557e91d55/details?tab=info\nSupposed you have installed docker, then you can pull the image by: \ndocker pull gcr.io/kaggle-gpu-images/python:v133\nBut it is very big, taking 50.2GB.", "2369551": "I just posted a new question, and maybe should have replied here first, so I'll do both. My new question: https://www.kaggle.com/competitions/asl-fingerspelling/discussion/428543\nbut essentially it's that I tried to do what's here, but I'm using a Mac with M1 chip and I don't think it will work because it's not an Nvidia GPU? If I'm wrong, and it's possible, and anyone wants to help me figure out how to do it correctly, I'm all ears. Thank you", "2340234": "Hey everyone! \n\nFirslty thanks @Zane for showing how to get the docker images that we can use. I also managed to get a functional environment going (or at least functional enough to run the introductory notebook). The trick is to use python venv instead of conda as there are a few packages that aren't available in conda's channels (most importantly mediapipe), so you would have to use pip to install them anyway. I am not quite sure why this is a problem, but it seems to cause issues with resolving versioning, leading to a lot of seeming conflicts. \n\nI attached my requirements.txt file here if anyone wants to recreate for themselves, not that since I am running on Apple silicon, I also installed tensorflow-metal (and tf-macos but that was installed on it's own by a separate wheel). \n\nHope this helps anyone else trying to run on their personal machine!", "2334527": "" } }