I want to use matterplot's MaskRCNN library: https://github.com/matterport/Mask_RCNN. This uses tensorflow (v1.x) implementation of MaskRCNN where the kernels have upgraded Tensorflow (v2.0). So I'm trying to first downgrade the TF version and the checking whether the GPU is available or not using tensorflow.test.is_gpu_available(), but this is returning False. Here is my kernel just for checking this: https://www.kaggle.com/kaushal2896/tensorflow-version-test
However when I do the same process on Google Colab, it's working fine. Here is my Colab notebook: https://colab.research.google.com/drive/1ecBB8Zzc4H0EXCZJ79dbjObgdXpHAT1H#scrollTo=PSTHIweDZzvS&line=1&uniqifier=1
\n\nWhat am I missing here? Why there is different behaviour?
", "messages": [ { "id": "805183", "postDate": "04/12/2020 12:49:21", "content": "I want to use matterplot's MaskRCNN library: https://github.com/matterport/Mask_RCNN. This uses tensorflow (v1.x) implementation of MaskRCNN where the kernels have upgraded Tensorflow (v2.0). So I'm trying to first downgrade the TF version and the checking whether the GPU is available or not using tensorflow.test.is_gpu_available(), but this is returning False. Here is my kernel just for checking this: https://www.kaggle.com/kaushal2896/tensorflow-version-test
However when I do the same process on Google Colab, it's working fine. Here is my Colab notebook: https://colab.research.google.com/drive/1ecBB8Zzc4H0EXCZJ79dbjObgdXpHAT1H#scrollTo=PSTHIweDZzvS&line=1&uniqifier=1
\n\nWhat am I missing here? Why there is different behaviour?
", "rawMarkdown": "I want to use matterplot's MaskRCNN library: https://github.com/matterport/Mask_RCNN. This uses tensorflow (v1.x) implementation of MaskRCNN where the kernels have upgraded Tensorflow (v2.0). So I'm trying to first downgrade the TF version and the checking whether the GPU is available or not using `tensorflow.test.is_gpu_available()`, but this is returning False. Here is my kernel just for checking this: https://www.kaggle.com/kaushal2896/tensorflow-version-test\n\nHowever when I do the same process on Google Colab, it's working fine. Here is my Colab notebook: https://colab.research.google.com/drive/1ecBB8Zzc4H0EXCZJ79dbjObgdXpHAT1H#scrollTo=PSTHIweDZzvS&line=1&uniqifier=1\n\nWhat am I missing here? Why there is different behaviour?", "votes": null }, { "id": "812817", "postDate": "04/19/2020 05:59:45", "content": "i meet the same problem , by the way , colab is free?
", "rawMarkdown": "i meet the same problem , by the way , colab is free?", "votes": null }, { "id": "813146", "postDate": "04/19/2020 12:22:08", "content": "Yes, colab is Free with P100 and K80 GPUs! Let me know if you find the solution of this problem.
", "rawMarkdown": "Yes, colab is Free with P100 and K80 GPUs! Let me know if you find the solution of this problem.", "votes": null }, { "id": "816349", "postDate": "04/22/2020 09:48:59", "content": "I guess:\nKaggle notebook doesn't have CUDA10.0 , but tf1.15.0 requires CUDA10.0 , maybe this's the reason.
\n\nColab has CUDA10.0 and maybe there's some softlink , so it's working fine.
", "rawMarkdown": "I guess:\nKaggle notebook doesn't have CUDA10.0 , but tf1.15.0 requires CUDA10.0 , maybe this's the reason. \n\nColab has CUDA10.0 and maybe there's some softlink , so it's working fine.", "votes": null }, { "id": "825505", "postDate": "04/29/2020 03:45:21", "content": "https://colab.research.google.com/drive/1p-Wc5WHjoC2W3js89cvB-nN6VHM3DsN0\nmy colab doesn't work , very confused
", "rawMarkdown": "https://colab.research.google.com/drive/1p-Wc5WHjoC2W3js89cvB-nN6VHM3DsN0\nmy colab doesn't work , very confused", "votes": null }, { "id": "832526", "postDate": "05/04/2020 07:59:55", "content": "!conda install tensorflow-gpu==1.14.0 -y \nhi , it can solve this problem
", "rawMarkdown": "!conda install tensorflow-gpu==1.14.0 -y \nhi , it can solve this problem", "votes": null } ], "comments": [ { "id": 812817, "author_name": "fanhaobei", "author_url": "", "post_date": "04/19/2020 05:59:45", "content": "i meet the same problem , by the way , colab is free?
", "votes": null, "replies": [ { "id": 813146, "author_name": "kaushal2896", "author_url": "", "post_date": "04/19/2020 12:22:08", "content": "Yes, colab is Free with P100 and K80 GPUs! Let me know if you find the solution of this problem.
", "votes": null, "replies": [] }, { "id": 816349, "author_name": "fanhaobei", "author_url": "", "post_date": "04/22/2020 09:48:59", "content": "I guess:\nKaggle notebook doesn't have CUDA10.0 , but tf1.15.0 requires CUDA10.0 , maybe this's the reason.
\n\nColab has CUDA10.0 and maybe there's some softlink , so it's working fine.
", "votes": null, "replies": [] } ] }, { "id": 825505, "author_name": "fanhaobei", "author_url": "", "post_date": "04/29/2020 03:45:21", "content": "https://colab.research.google.com/drive/1p-Wc5WHjoC2W3js89cvB-nN6VHM3DsN0\nmy colab doesn't work , very confused
", "votes": null, "replies": [] }, { "id": 832526, "author_name": "fanhaobei", "author_url": "", "post_date": "05/04/2020 07:59:55", "content": "!conda install tensorflow-gpu==1.14.0 -y \nhi , it can solve this problem
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "805183": "I want to use matterplot's MaskRCNN library: https://github.com/matterport/Mask_RCNN. This uses tensorflow (v1.x) implementation of MaskRCNN where the kernels have upgraded Tensorflow (v2.0). So I'm trying to first downgrade the TF version and the checking whether the GPU is available or not using `tensorflow.test.is_gpu_available()`, but this is returning False. Here is my kernel just for checking this: https://www.kaggle.com/kaushal2896/tensorflow-version-test\n\nHowever when I do the same process on Google Colab, it's working fine. Here is my Colab notebook: https://colab.research.google.com/drive/1ecBB8Zzc4H0EXCZJ79dbjObgdXpHAT1H#scrollTo=PSTHIweDZzvS&line=1&uniqifier=1\n\nWhat am I missing here? Why there is different behaviour?", "812817": "i meet the same problem , by the way , colab is free?", "813146": "Yes, colab is Free with P100 and K80 GPUs! Let me know if you find the solution of this problem.", "816349": "I guess:\nKaggle notebook doesn't have CUDA10.0 , but tf1.15.0 requires CUDA10.0 , maybe this's the reason. \n\nColab has CUDA10.0 and maybe there's some softlink , so it's working fine.", "825505": "https://colab.research.google.com/drive/1p-Wc5WHjoC2W3js89cvB-nN6VHM3DsN0\nmy colab doesn't work , very confused", "832526": "!conda install tensorflow-gpu==1.14.0 -y \nhi , it can solve this problem" }, "source": "meta" }