Anyone is running deep learning model on IMac ? How to use Tensorflow, as it seems not supported from Mac? Also, do we need to use the external grpahics or the inbuild Graphics card is sufficient. I have looked in the Internet, however din't find suitable/optimized way to more foreward of whether to buy external GPU ?
\n\nThanks a ton in advance
", "messages": [ { "id": "432084", "postDate": "12/03/2018 11:37:58", "content": "Anyone is running deep learning model on IMac ? How to use Tensorflow, as it seems not supported from Mac? Also, do we need to use the external grpahics or the inbuild Graphics card is sufficient. I have looked in the Internet, however din't find suitable/optimized way to more foreward of whether to buy external GPU ?
\n\nThanks a ton in advance
", "rawMarkdown": "Anyone is running deep learning model on IMac ? How to use Tensorflow, as it seems not supported from Mac? Also, do we need to use the external grpahics or the inbuild Graphics card is sufficient. I have looked in the Internet, however din't find suitable/optimized way to more foreward of whether to buy external GPU ?\n\nThanks a ton in advance", "votes": null }, { "id": "432143", "postDate": "12/03/2018 13:10:31", "content": "Hello MLS,
\n\nLook at this tutorial to run on Imac : @mattias.arro/installing-tensorflow-1-2-from-sources-with-gpu-support-on-macos-4f2c5cab8186\">https://medium.com/@mattias.arro/installing-tensorflow-1-2-from-sources-with-gpu-support-on-macos-4f2c5cab8186
\n\nUsually, the better your GPU is, the better are the performance.. what GPU do you have ?
", "rawMarkdown": "Hello MLS,\n\nLook at this tutorial to run on Imac : https://medium.com/@mattias.arro/installing-tensorflow-1-2-from-sources-with-gpu-support-on-macos-4f2c5cab8186\n\nUsually, the better your GPU is, the better are the performance.. what GPU do you have ?", "votes": null }, { "id": "432246", "postDate": "12/03/2018 15:49:16", "content": "AMD support is limited at best. Mac support is even worse. You could consider dual booting with Linux, and you'd still have to resolve the AMD backend issues. I tried to setup in OSX years ago, but I never got it working. Worst case, you could prototype in the cloud on a PaperSpace (or your favorite provider) instance.
", "rawMarkdown": "AMD support is limited at best. Mac support is even worse. You could consider dual booting with Linux, and you'd still have to resolve the AMD backend issues. I tried to setup in OSX years ago, but I never got it working. Worst case, you could prototype in the cloud on a PaperSpace (or your favorite provider) instance.", "votes": null }, { "id": "433068", "postDate": "12/04/2018 16:32:09", "content": "Thanks Simeon, Assous
\n\nHi Assous,
\n\nI have not yet bought external GPU in the fact that, if it will not configured/worked. Currently my Imac - 2017 has \"Radeon Pro Vega 56 8176 MB\", i though i would give a try, however it seems the tensor flow doesn't support on the AMD \nRocm (raised the question in stack overflow link)
\n\nNow, the only option left
\n (1) I need to go for the external GPU, as current Internal GPU is not able to support the Tensor Flow, as it has AMD Rocm
As going through to the Cloud option is costlier and long term is not a viable option
\n\nThanks.
", "rawMarkdown": "Thanks **Simeon**, **Assous**\n\nHi **Assous**,\n\n I have not yet bought external GPU in the fact that, if it will not configured/worked. Currently my Imac - 2017 has \"Radeon Pro Vega 56 8176 MB\", i though i would give a try, however it seems the tensor flow doesn't support on the AMD \nRocm (raised the question in stack overflow [link][1]) \n\n Now, the only option left \n (1) I need to go for the external GPU, as current Internal GPU is not able to support the Tensor Flow, as it has AMD Rocm\n \n As going through to the Cloud option is costlier and long term is not a viable option\n\nThanks.\n\n\n [1]: https://stackoverflow.com/questions/52871561/how-to-install-amd-rocm-in-apple-imac-pro-2017-to-further-install-tensorflow-fo", "votes": null }, { "id": "433203", "postDate": "12/04/2018 20:43:08", "content": "No, you can’t do reasonable deep learning on iMac, at least not image ones, which have a lot of parameters. I managed to do some cpu-only nlp text-model (charnn) running 200k epochs on a 1MB dataset overnight. I bet this would have taken 10 times less on a gpu.\nIf budget is a problem, you can do cloud based deep learning for free either on Kaggle kernels or on Google Colab. On the later you can pick the kind of machine(gpu or cpu) you want your code to run on. There are also other cheap options with coupon codes for multiple hours of cloud based gpu computing, just keep looking on the deep learning forums or slack channels.\nSo there’s really no reason not to get started, just go for it.
", "rawMarkdown": "No, you can’t do reasonable deep learning on iMac, at least not image ones, which have a lot of parameters. I managed to do some cpu-only nlp text-model (charnn) running 200k epochs on a 1MB dataset overnight. I bet this would have taken 10 times less on a gpu.\nIf budget is a problem, you can do cloud based deep learning for free either on Kaggle kernels or on Google Colab. On the later you can pick the kind of machine(gpu or cpu) you want your code to run on. There are also other cheap options with coupon codes for multiple hours of cloud based gpu computing, just keep looking on the deep learning forums or slack channels.\nSo there’s really no reason not to get started, just go for it.", "votes": null }, { "id": "433709", "postDate": "12/05/2018 11:37:23", "content": "If you choose to use cloud based GPU(s), look at collab or kaggle kernels. \nAlso, GCP offers 300$ as a sign-up bonus, which might help.
\n\nPS: The frustration of setting everything up >> Using a Cloud based solution (Based on my experience)
", "rawMarkdown": "If you choose to use cloud based GPU(s), look at collab or kaggle kernels. \nAlso, GCP offers 300$ as a sign-up bonus, which might help. \n\nPS: The frustration of setting everything up >> Using a Cloud based solution (Based on my experience)", "votes": null }, { "id": "434332", "postDate": "12/06/2018 08:06:43", "content": "Thanks Alexandru, init27.
\n\nHowever would go with the physical EGPU setup wth IMac or without Imac (Windows setup), will not thinking of going to cloud based GPU as its costlier if i run 24X7 for some days. Also long term serious deep learning project would be favorable for physical EGPU.
\n\nAnyone has physical EGPU for their Mac ? Help me the Setup end-to-end (Link / Document) ?
\n\nThanks.
", "rawMarkdown": "Thanks Alexandru, init27.\n\nHowever would go with the physical EGPU setup wth IMac or without Imac (Windows setup), will not thinking of going to cloud based GPU as its costlier if i run 24X7 for some days. Also long term serious deep learning project would be favorable for physical EGPU.\n\nAnyone has physical EGPU for their Mac ? Help me the Setup end-to-end (Link / Document) ?\n\nThanks.", "votes": null }, { "id": "435196", "postDate": "12/07/2018 16:44:40", "content": "I thought this might be of interest to those on a budget: https://medium.com/swlh/how-to-train-keras-model-x20-times-faster-with-tpu-for-free-cac6cf5089cb
", "rawMarkdown": "I thought this might be of interest to those on a budget: https://medium.com/swlh/how-to-train-keras-model-x20-times-faster-with-tpu-for-free-cac6cf5089cb", "votes": null } ], "comments": [ { "id": 432143, "author_name": "xtelima", "author_url": "", "post_date": "12/03/2018 13:10:31", "content": "Hello MLS,
\n\nLook at this tutorial to run on Imac : @mattias.arro/installing-tensorflow-1-2-from-sources-with-gpu-support-on-macos-4f2c5cab8186\">https://medium.com/@mattias.arro/installing-tensorflow-1-2-from-sources-with-gpu-support-on-macos-4f2c5cab8186
\n\nUsually, the better your GPU is, the better are the performance.. what GPU do you have ?
", "votes": null, "replies": [] }, { "id": 432246, "author_name": "badtyprr", "author_url": "", "post_date": "12/03/2018 15:49:16", "content": "AMD support is limited at best. Mac support is even worse. You could consider dual booting with Linux, and you'd still have to resolve the AMD backend issues. I tried to setup in OSX years ago, but I never got it working. Worst case, you could prototype in the cloud on a PaperSpace (or your favorite provider) instance.
", "votes": null, "replies": [] }, { "id": 433068, "author_name": "sushanta", "author_url": "", "post_date": "12/04/2018 16:32:09", "content": "Thanks Simeon, Assous
\n\nHi Assous,
\n\nI have not yet bought external GPU in the fact that, if it will not configured/worked. Currently my Imac - 2017 has \"Radeon Pro Vega 56 8176 MB\", i though i would give a try, however it seems the tensor flow doesn't support on the AMD \nRocm (raised the question in stack overflow link)
\n\nNow, the only option left
\n (1) I need to go for the external GPU, as current Internal GPU is not able to support the Tensor Flow, as it has AMD Rocm
As going through to the Cloud option is costlier and long term is not a viable option
\n\nThanks.
", "votes": null, "replies": [] }, { "id": 433203, "author_name": "alexbrie", "author_url": "", "post_date": "12/04/2018 20:43:08", "content": "No, you can’t do reasonable deep learning on iMac, at least not image ones, which have a lot of parameters. I managed to do some cpu-only nlp text-model (charnn) running 200k epochs on a 1MB dataset overnight. I bet this would have taken 10 times less on a gpu.\nIf budget is a problem, you can do cloud based deep learning for free either on Kaggle kernels or on Google Colab. On the later you can pick the kind of machine(gpu or cpu) you want your code to run on. There are also other cheap options with coupon codes for multiple hours of cloud based gpu computing, just keep looking on the deep learning forums or slack channels.\nSo there’s really no reason not to get started, just go for it.
", "votes": null, "replies": [ { "id": 435196, "author_name": "badtyprr", "author_url": "", "post_date": "12/07/2018 16:44:40", "content": "I thought this might be of interest to those on a budget: https://medium.com/swlh/how-to-train-keras-model-x20-times-faster-with-tpu-for-free-cac6cf5089cb
", "votes": null, "replies": [] } ] }, { "id": 433709, "author_name": "init27", "author_url": "", "post_date": "12/05/2018 11:37:23", "content": "If you choose to use cloud based GPU(s), look at collab or kaggle kernels. \nAlso, GCP offers 300$ as a sign-up bonus, which might help.
\n\nPS: The frustration of setting everything up >> Using a Cloud based solution (Based on my experience)
", "votes": null, "replies": [] }, { "id": 434332, "author_name": "sushanta", "author_url": "", "post_date": "12/06/2018 08:06:43", "content": "Thanks Alexandru, init27.
\n\nHowever would go with the physical EGPU setup wth IMac or without Imac (Windows setup), will not thinking of going to cloud based GPU as its costlier if i run 24X7 for some days. Also long term serious deep learning project would be favorable for physical EGPU.
\n\nAnyone has physical EGPU for their Mac ? Help me the Setup end-to-end (Link / Document) ?
\n\nThanks.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "432084": "Anyone is running deep learning model on IMac ? How to use Tensorflow, as it seems not supported from Mac? Also, do we need to use the external grpahics or the inbuild Graphics card is sufficient. I have looked in the Internet, however din't find suitable/optimized way to more foreward of whether to buy external GPU ?\n\nThanks a ton in advance", "432143": "Hello MLS,\n\nLook at this tutorial to run on Imac : https://medium.com/@mattias.arro/installing-tensorflow-1-2-from-sources-with-gpu-support-on-macos-4f2c5cab8186\n\nUsually, the better your GPU is, the better are the performance.. what GPU do you have ?", "432246": "AMD support is limited at best. Mac support is even worse. You could consider dual booting with Linux, and you'd still have to resolve the AMD backend issues. I tried to setup in OSX years ago, but I never got it working. Worst case, you could prototype in the cloud on a PaperSpace (or your favorite provider) instance.", "433068": "Thanks **Simeon**, **Assous**\n\nHi **Assous**,\n\n I have not yet bought external GPU in the fact that, if it will not configured/worked. Currently my Imac - 2017 has \"Radeon Pro Vega 56 8176 MB\", i though i would give a try, however it seems the tensor flow doesn't support on the AMD \nRocm (raised the question in stack overflow [link][1]) \n\n Now, the only option left \n (1) I need to go for the external GPU, as current Internal GPU is not able to support the Tensor Flow, as it has AMD Rocm\n \n As going through to the Cloud option is costlier and long term is not a viable option\n\nThanks.\n\n\n [1]: https://stackoverflow.com/questions/52871561/how-to-install-amd-rocm-in-apple-imac-pro-2017-to-further-install-tensorflow-fo", "433203": "No, you can’t do reasonable deep learning on iMac, at least not image ones, which have a lot of parameters. I managed to do some cpu-only nlp text-model (charnn) running 200k epochs on a 1MB dataset overnight. I bet this would have taken 10 times less on a gpu.\nIf budget is a problem, you can do cloud based deep learning for free either on Kaggle kernels or on Google Colab. On the later you can pick the kind of machine(gpu or cpu) you want your code to run on. There are also other cheap options with coupon codes for multiple hours of cloud based gpu computing, just keep looking on the deep learning forums or slack channels.\nSo there’s really no reason not to get started, just go for it.", "433709": "If you choose to use cloud based GPU(s), look at collab or kaggle kernels. \nAlso, GCP offers 300$ as a sign-up bonus, which might help. \n\nPS: The frustration of setting everything up >> Using a Cloud based solution (Based on my experience)", "434332": "Thanks Alexandru, init27.\n\nHowever would go with the physical EGPU setup wth IMac or without Imac (Windows setup), will not thinking of going to cloud based GPU as its costlier if i run 24X7 for some days. Also long term serious deep learning project would be favorable for physical EGPU.\n\nAnyone has physical EGPU for their Mac ? Help me the Setup end-to-end (Link / Document) ?\n\nThanks.", "435196": "I thought this might be of interest to those on a budget: https://medium.com/swlh/how-to-train-keras-model-x20-times-faster-with-tpu-for-free-cac6cf5089cb" }, "source": "meta" }