This is the best environment setup tutorial I've ever seen. I just got my personal MacBook Air M1 ~30 min ago and I'm already up and running with TF on the GPU. Well done, sir!
@@mrdbourke when I try to use my apple gpu with pytorch on jupyter notebook, it said that my kernel died. How can I make my kernel works without this thing happen? PS: I try to use a transformer model to train on an algorithm to detect things on images
Thank you Daniel! I bought the M1 MBP so that I could learn and experiment. Successfully installed and tested TensorFlow-metal. Enrolled in your TemsorFlow Developer Certificate course. Great course, great explanations of concepts and exercises!
WOW!! Speechless literally I was just about to give up and start doing Machine Learning on my old pc if it were not for this I would have been coding on a toaster. Thank you Daniel very though rough video and very helpful.
You are a life saver man! Had been stuck trying to get tensor flow working on my m1(I'm new to Python and ML) and after trying out countless ways I finally found your video. Thanks a ton!
Still waiting on my M1 Max to show up after ordering the week after the release. Thanks for these videos. Would love to see more about how you setup you dev environment in general.
Woah! They must be popular if you still don’t have it… as for my dev setup I usually make a Conda environment (similar to the video) for each different project. The environment lives inside the project folder so everything for a project is in one place.
@@mrdbourke what are your thoughts on using containers to isolate each development tool from the host computer? I have been using containerized Python, pip, pipenv, pylint, and Jupyter to (1) ensure the same development environment is easily reproducible and (2) avoid "polluting" the host computer with globally-installed development tools and libraries.
thanks a lot for this video, i tried many diff videos on yt and yet none of them worked but i successfully installed tf using ur video, really glad i fond this video
Hey man. You saved my life. I am very grateful for your help. I ended up using my old windows laptop for my project because tensorflow was not running on my m1.
It is an incredible video as it is the first time to successfully download tensorflow libraries on my mac. Unfortunately, the step when I am supposed to successfully run the code on Jupyter notebook did not work (1st create new python files leads to: "500: Internal Server")(2nd if the 1st step worked using: "pip instal --upgarde nbconvert", then when executing the code libraries like numpy are not found). These libraries such as numpy, pandas, and matplotlib could be imported successfully if I opened the python files using visual studio code, except for tensorflow. I'm still looking for an explanation and a solution for such problem.
Thank you for this. I have to try this with my M1 MacMini. Bought it about 6 months ago and I was disappointed how hard it was to make it work with ML/DL libraries. M1 has been mainly collecting dust. Time to do upgrade to OSX and try your step by step video. Maybe there is still some problems, but month by month we get new libraries which support ARM better. (Now I do not remember which libraries were hard to get going on 6 months ago, maybe it was gym and ml-agents, mainly reinforcement stuff)
Hey Harri, I was the same when the M1 first came out. But now I find it works basically the same as all of the other machines I do machine learning on. Especially with TensorFlow-macos and TensorFlow-metal, installing these were a breeze to get TensorFlow running on the GPU
@@mrdbourke I just installed on a M1 Mac mini and it doesn't work. The import tensorflow as tf line in the notebook gives the error: dlopen ........ Symbol not found: _OBJC_CLASS_$_MPSGraphCompilationDescriptor
Hi Daniel, thank you for the Video, after following the instructions while connecting the Jupiter notebook I am getting error "RuntimeError Traceback (most recent call last) RuntimeError: module compiled against API version 0xf but this version of numpy is 0xe --------------------------------------------------------------------------- ImportError Traceback (most recent call last) ImportError: numpy.core.multiarray failed to import The above exception was the direct cause of the following exception: SystemError Traceback (most recent call last) SystemError: returned a result with an error set The above exception was the direct cause of the following exception: " Request you please advise
I completed all the steps until the "jupyter notebook" script. However, when I created and entered a new python notebook "500: Internal Server Error" appeared. What could be the reason and how can I fix this?
I have an M3 Max and the metal install fails. Is there an updated process? I get the following python -m pip install tensorflow-metal ERROR: Could not find a version that satisfies the requirement tensorflow-metal (from versions: none) ERROR: No matching distribution found for tensorflow-metal
If I follow these steps and open up jupyter notebook from the terminal it works but then creating a new file in this directory throws a 500: INTERNAL SERVER ERROR message. If I try opening jupyter notebook from my anaconda environment it works and I can create a new file but when I try to import tensorflow it tells me he doesn't know that module. Anyone else got a similar problem and can help me out? Thank you in advance
Hi Daniel, thanks for this comprehensive tutorial. I got some questions about creating an environment: 1. Is it a good practice to create an environment on specific directory for every project? Or we create different environments on the same directory like /Users/username/. 2. If I install different environment on different specific directories, do I have to go install again all Tensorflow dependencies for every single one of them? Thanks.
Hey Hoang, you’re welcome legend! 1. I personally usually create an environment for every project. Or I setup one universal environment for playing around with TensorFlow or something and do lots of projects inside that universal environment. 2. Usually with every new environment you install the dependencies you’d like to use in it. Perhaps there might be a way to link an existing install of TensorFlow to that new environment but I’m not 100% sure on this.
Hi Daniel, I recently purchased my Macbook Air M1, I tried running your process and it got executed successfully. At the last I have below given results, please suggest what needs to be corrected here:- --------------------------------------------------------------------------- ModuleNotFoundError Traceback (most recent call last) Input In [1], in () 2 import pandas as pd 3 import sklearn ----> 4 import tensorflow as tf 5 import matplotlib.pyplot as plt 7 # Check for TensorFlow GPU access ModuleNotFoundError: No module named 'tensorflow' Please help, Thanks in Advance :)
Thanks for the awesome tutorial! Just a few questions, if you don't mind: 1. Do I need to activate/deactivate the env every time I want to work on my ML/DL projects? 2. When I want to open Jupyter notebook to work on my existing ML/DL project, do I need to do so by activating the env inside the tensorflow_test folder every time? 3. Is is good practice to have multiple envs for multiple projects? Will it 'waste' (because of redundant packages) storage on my Mac? 4. Will this work with VS Code? I wanted to try out the GitHub Copilot feature... 5. If I need to install a package, do I just pip install in the home directory, or first activate the env and then pip install the package? 6. Is there is a reason why you chose Homebrew over Macports? Again, thanks a lot for the much needed tutorial!
1. yes everytime you open the terminal you usually start in your base env 2. the env has nothing to do with the folder, after you activate the env you can change your directory 3.yes it best practice, because you may need different packages version for different projects, of course there is redundant, but less problems 4. - 5. first activate the env you want it to be installed the do the rest, you can use "conda install pandas" or something like this as well 6.- you could have also asked google
Yo yo! You're welcome. I'll answer your questions. 1. Yes, for specific environments + projects you'll need to activate the environment to continue working on the project. You don't necessarily need to deactivate it (if you're going to come back to it). 2. Yes. If you want to work on a Jupyter Notebook within the project and have access to the packages there, you should activate the environment. 3. Here's what I do: 1x large environment for hacking around/hobby projects + 1x environment for every new larger project. If I don't use an environment in a few months, I delete it (but I also export the packages to a YAML file). Google "conda export environment packages to YAML". 4. Yes it will. Activate the environment and project folder through VS Code's terminal. 5. If you want to install the package in the environment, the environment should be active. If you just want to install the package globally, you can run pip install without and activate environment. 6. I have never heard of Macports (this isn't saying anything against it). But I have used Homebrew for many years and have the most experience with it. Let me know if you have any follow ups.
@@mrdbourke Thanks for the detailed response! I am waiting on my Macbook Pro 14 inch base model. Was hopeful for a Christmas delivery, but looks like it'll be delivered around mid-January. Regardless, I will be following your tutorial and then follow up on the 10 hour stream. Thanks a lot!
@@kobold2376 Thanks for the response. Just got back after a hiatus and saw your answer. True, I could have just Googled everything, and I am sure I will when I am setting up my Macbook when it is finally delivered. Anyway, thanks!
@@mrdbourke pipenv has greatly streamlined the process of creating and activating virtual environments as well as installing Python libraries. within a project's virtual environment.
ModuleNotFoundError: No module named 'tensorflow' can you please reply. while i am running the code import numpy as np import pandas as pd import sklearn import tensorflow as tf import matplotlib.pyplot as plt # Check for TensorFlow GPU access print(f"TensorFlow has access to the following devices: {tf.config.list_physical_devices()}") # See TensorFlow version print(f"TensorFlow version: {tf.__version__}")
First: thank you for the video, second: the Mac and M1 is an adventure video-game, third: you need a lot of British humor to survive it. It is now the third try I do, following different you-tube clips which shows a) it is all very simple and b) if this doesn't work, try something else, maybe it will work. Mixed in-between are some University Teachers, which explain all very well, till you just get to something really complicate and they almost alway bridge this by saying; this is now simple and I need not to be explained further...So this where my last 3 months of sufferings. Well I tried it now with your video, I learned a lot, but when I get to install Miniforge it arrives after a copyright notice on conda-forge to. "Redistribution and use in source and binary forms, with or without: " and goes into an infinite loop. I returned, reinstalled Brew and everything else, but no way out. Any idea what I could do now?
Can We change the Name of the folder from “tensorflow-test” to by renaming that folder (any number of times I want ) , after doing all the process successfully? Does it create any conflict? Thank you Sir for giving such a nice and simple explanation
Yes you can, you can name the folder however you want. Though if you create it as “tensorflow-test” and setup the environment with “conda create -prefix ./env” the env will rename at the original path (e.g. “tensorflow-test/env”) If you want to change the directory name *after* you made the env, you might have to look into renaming the Conda env path too
Great video for the quick setup. However, I think Mac M1 GPU is useless for ML development since it never works correctly. Actual CPU works much better. GPU is just there for the sake of it.
Hello Daniel. As a university student trying to dive into machine learning, setting up the environment (especially with an m1 mac) is the most draining thing to do. I wish I found this video earlier. This is the best video on this topic. Could you also make a tutorial for installing the TFOD API on m1 macs. I have scoured the internet and no one has an answer, resorted to using colab but that has limited capabilities unless you pay for it. Thanks again.
You can do that yes with "brew install miniforge", however I wanted to link to the Miniforge website in the tutorial so someone knows where it comes from.
In this way, for every project should i have to create a new directory and set environment? Can you please show how to set up a an universal environment and play around different projects
Thank you so much for making this video, it's exactly what I've been looking for. Just received my new macbook pro and can't wait to get it set up. :-) One question, if I want to use PyCharm as my IDE can I just select the virtual environment we have created in this video in PyCharm and TensorFlow will still use the GPU?
Thank you for your detail explanations. But I have a problem to use jupyter notebook. When I type "jupyter notebook" like you (11:15), occured. Someone says its reason is UTF-8 encoding, but i can't solve the problem though i changed the setting. If you know anything about this, could you tell me? Again, thanks a lot for the video!
At 11:15 my jupyter notebook was opening but not creating any new file. So I installed two more packages and the issue was gone : pip install chardet cchardet Now again run "jupyter notebook" and it will allow you to create new file. Cheers
thanks thanks thanks ! amazing ! just one question. How do I into to env in different way? I can not use " Conda activate env", only Conda activate Users/.../env
You have to make sure you’re in the right folder (where your env is). E.g. if your env is in “TensorFlow-test” folder you need to be in that folder to use “conda activate ./env”
Thanks a million Daniel. Though, I have one question to ask, Is there any reason that you downloaded Miniforge at the Home directory instead of the root directory? when I've compared the anaconda location with my other intel devices (where I used brew install) they just downloaded Under the root/opt directory and there seems to be a little difference in making new environments with the following common command "conda create --name env_name python==version"
Thank you Daniel! I have NO idea how any of this works and first time I've opened the terminal - and still don't understand any of it, but it works. How can I get the environment to work on my Spyder installation? In the preferences I can change the environment to "/Users/my_user/tensorflow-test/env/bin/python" and it works with python, but says I do not have TensorFlow installed. Sorry, but I've only ever worked with graphic interfaces so I have no idea what I'm doing here - but the tutorial was awesome 😊
you may have to make sure Spyder knows that you’re using “conda” as the environment - if you can activate an environment using “conda” in Spyder I’d do that. Perhaps search “use conda environment with Spyder” and then activate the environment we created in this video
its working pretty good but i have question i have exit the terminal after all this and then when i open the terminal and type jupyter notebook it said invalid command could you please help me with this ?
This is the best environment setup tutorial I've ever seen. I just got my personal MacBook Air M1 ~30 min ago and I'm already up and running with TF on the GPU. Well done, sir!
Thank you Daniel! Enjoy!
I wish I discovered this earlier, I wasted all day trying to setup a docker.
@@mrdbourke when I try to use my apple gpu with pytorch on jupyter notebook, it said that my kernel died. How can I make my kernel works without this thing happen?
PS: I try to use a transformer model to train on an algorithm to detect things on images
Can help me decide, is MacBook air m1 good for starting on the ai and ML filed , or better to get a Nvidia GPU laptop ?
Please reply to this.I'm buying m2 air now
Thank you very much. I was trying to install tensorflow for two days, i couldn't find any way, your video solved my problem. Thank you very very much.
Thanks Daniel! After trying multiple other tutorials I was going to give up, but I got TensorFlow working with this tutorial!
Thank you Daniel! I bought the M1 MBP so that I could learn and experiment. Successfully installed and tested TensorFlow-metal. Enrolled in your TemsorFlow Developer Certificate course. Great course, great explanations of concepts and exercises!
WOW!! Speechless literally I was just about to give up and start doing Machine Learning on my old pc if it were not for this I would have been coding on a toaster.
Thank you Daniel very though rough video and very helpful.
You are a life saver man! Had been stuck trying to get tensor flow working on my m1(I'm new to Python and ML) and after trying out countless ways I finally found your video. Thanks a ton!
Can we run code on gpu
There are dozens of videos that I have not been able to utilize, this one is so useful. Thank you!
Thank you! Glad you enjoyed!
Dear Sir，I am an international student at UOM，and I was doing preview for the 2nd year study，your video helped me a lot on tensorflow！！thank you！
You’re welcome Chen!
Still waiting on my M1 Max to show up after ordering the week after the release. Thanks for these videos. Would love to see more about how you setup you dev environment in general.
Woah! They must be popular if you still don’t have it… as for my dev setup I usually make a Conda environment (similar to the video) for each different project. The environment lives inside the project folder so everything for a project is in one place.
Into my third week:)
@@mrdbourke what are your thoughts on using containers to isolate each development tool from the host computer? I have been using containerized Python, pip, pipenv, pylint, and Jupyter to (1) ensure the same development environment is easily reproducible and (2) avoid "polluting" the host computer with globally-installed development tools and libraries.
amazing, I tried to get all this installed for the last few days and failed, with the video it was so easy. Thank you so much!
and now my M2Pro chip comes to life in tensorFlow. Thanks a million, man!!
You saved my £1500 man i was about to break my mac pro and wanted to get windows
You are such a beauty thank you very much everything worked on my mac
Enjoy!
Thanks! These instructions worked on M3 Pro as well! You saved my day!
Glad to hear it worked!
Literally the only guide on the whole internet that actually works
thanks a lot for this video, i tried many diff videos on yt and yet none of them worked but i successfully installed tf using ur video, really glad i fond this video
Thank you so much daniel....i have been trying from the past 4 months to install tensorflow on mac....this really helped me a lot :-))
You’re welcome Sai!
Thanks this worked better than the other solutions I've seen on UA-cam
gone through many tutorials but found this one the best....and works,, so, I am subscribing you...
Glad to hear it works!
Hey man. You saved my life. I am very grateful for your help. I ended up using my old windows laptop for my project because tensorflow was not running on my m1.
Best tutorial ever! Now I can finally use my mac to do the deep learning locally. Thank you so much!
Best video! The only video all things worked On M1 Mac as per 2022
I recently purchased macbook air. I was facing difficulty in installing ML libraries. This is very helpful. Thanks :)
Hey Pankaj! Glad to hear :) did you get it all setup?
BRO YOU JUST SAVED ME DAYS OF RESEARCH. THANK YOU SO MUCH
bought the mac pro 14 the same day you release this video. Thanks so much for this video
You’re welcome Rich! Enjoy your new computer
Daniel - sincere thanks for sharing your hard won knowledge and skills. You are an open and supportive person. Cheers ! ... Well Done !!
Thanks Daniel, this was very helpful. Can you also make a similar tutorial on setting up PyTorch on M1 Macs? That would be super helpful.
PyTorch GPU doesn’t work and will never work on M1, cpu version can be installed follow their getting started guide
Let me look into the best setup for PyTorch on M1!
@@mrdbourke Thanks man, I will be waiting for the PyTorch setup tutorial.
@@mrdbourke waiting for PyTorch and Transformers as well! I have troubles with those two as well.
@@trefoilz7355 Man I hope there is! , to avoid the disconnection on google colab.
Awesome Video Daniel !! Such a crisp explanation for the process
Only thing that worked for me after multiple hours!! Thank you so much!!!
Glad it worked Olivia! Happy machine learning!
Omg thanks ive struggled for so long to get this running 🏃
But you came to save my in the Mac terminal lost souls
It is an incredible video as it is the first time to successfully download tensorflow libraries on my mac. Unfortunately, the step when I am supposed to successfully run the code on Jupyter notebook did not work (1st create new python files leads to: "500: Internal Server")(2nd if the 1st step worked using: "pip instal --upgarde nbconvert", then when executing the code libraries like numpy are not found). These libraries such as numpy, pandas, and matplotlib could be imported successfully if I opened the python files using visual studio code, except for tensorflow. I'm still looking for an explanation and a solution for such problem.
As always great content sir... kudos !!! My new machine arrives this week, very excited to reproduce this tutorial on that.
Thank you Vlademir! Enjoy your new machine
The legend back again. Thank you Daniel!!
You’re welcome Ali! Plenty more videos to come :)
Thanks a million, after spending several hours trying to solve it, I finally managed through you. Cheers :)
You’re welcome Shady! Glad you got it working!
Thank you for this. I have to try this with my M1 MacMini. Bought it about 6 months ago and I was disappointed how hard it was to make it work with ML/DL libraries. M1 has been mainly collecting dust. Time to do upgrade to OSX and try your step by step video. Maybe there is still some problems, but month by month we get new libraries which support ARM better. (Now I do not remember which libraries were hard to get going on 6 months ago, maybe it was gym and ml-agents, mainly reinforcement stuff)
Hey Harri, I was the same when the M1 first came out. But now I find it works basically the same as all of the other machines I do machine learning on. Especially with TensorFlow-macos and TensorFlow-metal, installing these were a breeze to get TensorFlow running on the GPU
@@mrdbourke I just installed on a M1 Mac mini and it doesn't work. The import tensorflow as tf line in the notebook gives the error: dlopen ........ Symbol not found: _OBJC_CLASS_$_MPSGraphCompilationDescriptor
Well, i hVe learned alot from you. I have took your lectures from udemy.. i am really thankful for that. Stay blessed and keep it up.
Thank you Irfan! Keep learning legend
Hi Daniel, thank you for the Video, after following the instructions while connecting the Jupiter notebook I am getting error "RuntimeError Traceback (most recent call last)
RuntimeError: module compiled against API version 0xf but this version of numpy is 0xe
---------------------------------------------------------------------------
ImportError Traceback (most recent call last)
ImportError: numpy.core.multiarray failed to import
The above exception was the direct cause of the following exception:
SystemError Traceback (most recent call last)
SystemError: returned a result with an error set
The above exception was the direct cause of the following exception:
" Request you please advise
same to me.
Thank you so much Daniel. Very happy and grateful to you 😊👌🏼👍🏼
Really helpful !! Will stay tuned to more of your videos.
Thank you Amit!
wauuu excellent video,Thanks a Trillon Daniel,
blessings to you.
I completed all the steps until the "jupyter notebook" script. However, when I created and entered a new python notebook "500: Internal Server Error" appeared. What could be the reason and how can I fix this?
Same
Same here, on macbook m1
I have an M3 Max and the metal install fails. Is there an updated process? I get the following
python -m pip install tensorflow-metal
ERROR: Could not find a version that satisfies the requirement tensorflow-metal (from versions: none)
ERROR: No matching distribution found for tensorflow-metal
Hey , Can you make one more video regarding the installation with the updated instructions. I think there have been few changes in the last year.
Thanks so much Dan, you are the best. Life saver than a CIU
You’re welcome Russell!
i have done the whole setup, but at the end it shows "500 internal server error ".
please resolve my querry
hey Daniel, I m having problem in installing tensorflow-text please can u make a video on it
thanks a lot. it helped me to get my macbook successfully work
Stoked to hear Mohamed!
If I follow these steps and open up jupyter notebook from the terminal it works but then creating a new file in this directory throws a 500: INTERNAL SERVER ERROR message. If I try opening jupyter notebook from my anaconda environment it works and I can create a new file but when I try to import tensorflow it tells me he doesn't know that module. Anyone else got a similar problem and can help me out? Thank you in advance
me too
Thanks a lot! It worked on my Mac.
Thank you so much for the video. I love it
Many thanks for the helpful tutorial
Hi Daniel, thanks for this comprehensive tutorial. I got some questions about creating an environment:
1. Is it a good practice to create an environment on specific directory for every project? Or we create different environments on the same directory like /Users/username/.
2. If I install different environment on different specific directories, do I have to go install again all Tensorflow dependencies for every single one of them?
Thanks.
Hey Hoang, you’re welcome legend!
1. I personally usually create an environment for every project. Or I setup one universal environment for playing around with TensorFlow or something and do lots of projects inside that universal environment.
2. Usually with every new environment you install the dependencies you’d like to use in it. Perhaps there might be a way to link an existing install of TensorFlow to that new environment but I’m not 100% sure on this.
Hi Daniel, I recently purchased my Macbook Air M1, I tried running your process and it got executed successfully. At the last I have below given results, please suggest what needs to be corrected here:-
---------------------------------------------------------------------------
ModuleNotFoundError Traceback (most recent call last)
Input In [1], in ()
2 import pandas as pd
3 import sklearn
----> 4 import tensorflow as tf
5 import matplotlib.pyplot as plt
7 # Check for TensorFlow GPU access
ModuleNotFoundError: No module named 'tensorflow'
Please help, Thanks in Advance :)
did you find the solution to this?
same here
Thanks for the awesome tutorial!
Just a few questions, if you don't mind:
1. Do I need to activate/deactivate the env every time I want to work on my ML/DL projects?
2. When I want to open Jupyter notebook to work on my existing ML/DL project, do I need to do so by activating the env inside the tensorflow_test folder every time?
3. Is is good practice to have multiple envs for multiple projects? Will it 'waste' (because of redundant packages) storage on my Mac?
4. Will this work with VS Code? I wanted to try out the GitHub Copilot feature...
5. If I need to install a package, do I just pip install in the home directory, or first activate the env and then pip install the package?
6. Is there is a reason why you chose Homebrew over Macports?
Again, thanks a lot for the much needed tutorial!
1. yes everytime you open the terminal you usually start in your base env
2. the env has nothing to do with the folder, after you activate the env you can change your directory
3.yes it best practice, because you may need different packages version for different projects, of course there is redundant, but less problems
4. -
5. first activate the env you want it to be installed the do the rest, you can use "conda install pandas" or something like this as well
6.-
you could have also asked google
Yo yo! You're welcome. I'll answer your questions.
1. Yes, for specific environments + projects you'll need to activate the environment to continue working on the project. You don't necessarily need to deactivate it (if you're going to come back to it).
2. Yes. If you want to work on a Jupyter Notebook within the project and have access to the packages there, you should activate the environment.
3. Here's what I do: 1x large environment for hacking around/hobby projects + 1x environment for every new larger project. If I don't use an environment in a few months, I delete it (but I also export the packages to a YAML file). Google "conda export environment packages to YAML".
4. Yes it will. Activate the environment and project folder through VS Code's terminal.
5. If you want to install the package in the environment, the environment should be active. If you just want to install the package globally, you can run pip install without and activate environment.
6. I have never heard of Macports (this isn't saying anything against it). But I have used Homebrew for many years and have the most experience with it.
Let me know if you have any follow ups.
@@mrdbourke Thanks for the detailed response! I am waiting on my Macbook Pro 14 inch base model. Was hopeful for a Christmas delivery, but looks like it'll be delivered around mid-January. Regardless, I will be following your tutorial and then follow up on the 10 hour stream. Thanks a lot!
@@kobold2376 Thanks for the response. Just got back after a hiatus and saw your answer.
True, I could have just Googled everything, and I am sure I will when I am setting up my Macbook when it is finally delivered.
Anyway, thanks!
@@mrdbourke pipenv has greatly streamlined the process of creating and activating virtual environments as well as installing Python libraries. within a project's virtual environment.
Seriously appreciate this video! Thank you
You’re welcome Michael! Happy machine learning :)
Great tutorial, however I would not limit the virtual env within a folder, the better option would be to install a global virtual env using conda.
Such a helpful video. Can't thank you enough.
Glad you enjoyed legend!
ModuleNotFoundError: No module named 'tensorflow'
can you please reply. while i am running the code
import numpy as np
import pandas as pd
import sklearn
import tensorflow as tf
import matplotlib.pyplot as plt
# Check for TensorFlow GPU access
print(f"TensorFlow has access to the following devices:
{tf.config.list_physical_devices()}")
# See TensorFlow version
print(f"TensorFlow version: {tf.__version__}")
First: thank you for the video, second: the Mac and M1 is an adventure video-game, third: you need a lot of British humor to survive it. It is now the third try I do, following different you-tube clips which shows a) it is all very simple and b) if this doesn't work, try something else, maybe it will work. Mixed in-between are some University Teachers, which explain all very well, till you just get to something really complicate and they almost alway bridge this by saying; this is now simple and I need not to be explained further...So this where my last 3 months of sufferings.
Well I tried it now with your video, I learned a lot, but when I get to install Miniforge it arrives after a copyright notice on conda-forge to. "Redistribution and use in source and binary forms, with or without: " and goes into an infinite loop.
I returned, reinstalled Brew and everything else, but no way out. Any idea what I could do now?
Basically wonderful tutorial Thanks a lot!
Can We change the Name of the folder from “tensorflow-test” to by renaming that folder (any number of times I want ) , after doing all the process successfully?
Does it create any conflict?
Thank you Sir for giving such a nice and simple explanation
Yes you can, you can name the folder however you want. Though if you create it as “tensorflow-test” and setup the environment with “conda create -prefix ./env” the env will rename at the original path (e.g. “tensorflow-test/env”)
If you want to change the directory name *after* you made the env, you might have to look into renaming the Conda env path too
Worked for me, i even run a computer vision code for segmenting roads in video files. Just had to ' conda install opencv'
Very nice video. Is possible to make a video on Windows Installation
@Daniel Bourke. Excellent tutorial. It worked perfectly. I'm using R for some CNN. Can you make a video on the TensorFlow, Keras, and R integration?
excellent job - helping a lot 🙂
many thanks for your video, I have a question, how can we reload conda after we reboot the mac.
Great video for the quick setup. However, I think Mac M1 GPU is useless for ML development since it never works correctly. Actual CPU works much better. GPU is just there for the sake of it.
Hello Daniel. As a university student trying to dive into machine learning, setting up the environment (especially with an m1 mac) is the most draining thing to do. I wish I found this video earlier. This is the best video on this topic.
Could you also make a tutorial for installing the TFOD API on m1 macs. I have scoured the internet and no one has an answer, resorted to using colab but that has limited capabilities unless you pay for it. Thanks again.
Glad you got it fixed and sorted! Thank you for the kind words
@@mrdbourke Anything on the TFOD API for M1 macs?
Great and clear instructions !! Thank you a lot. :-))
Great content! Why not installing miniforge using brew?
You can do that yes with "brew install miniforge", however I wanted to link to the Miniforge website in the tutorial so someone knows where it comes from.
@@mrdbourke, thanks. That's how I did it, I was wondering if maybe it would cause some package compatibility issues on M1
The tutorial is clear and nice -- but just that I am getting the error on my M1 iMac.
What error? You can post it on the issues page of the GitHub repo for help
@@mrdbourke OK -- i just posted the error message there. Thanks
life saver. Thank you
This would have saved me a lot of time. I eventually got it set up.
Good to hear!
In this way, for every project should i have to create a new directory and set environment? Can you please show how to set up a an universal environment and play around different projects
Everything worked but when i try to create a new python file in jupyter as shown, it says "500 internal server error" how to fix this?
Thank you so much for making this video, it's exactly what I've been looking for. Just received my new macbook pro and can't wait to get it set up. :-)
One question, if I want to use PyCharm as my IDE can I just select the virtual environment we have created in this video in PyCharm and TensorFlow will still use the GPU?
You're welcome Sebastian! How did the setup go? And yes, you can still use a GPU in PyCharm with an environment active
Thank you for your detail explanations.
But I have a problem to use jupyter notebook. When I type "jupyter notebook" like you (11:15), occured.
Someone says its reason is UTF-8 encoding, but i can't solve the problem though i changed the setting.
If you know anything about this, could you tell me?
Again, thanks a lot for the video!
THANKS! GREAT VIDEO
At 11:15 my jupyter notebook was opening but not creating any new file. So I installed two more packages and the issue was gone : pip install chardet cchardet
Now again run "jupyter notebook" and it will allow you to create new file. Cheers
Thanks so much Daniel. I was waiting to use my M1 mac air. Do you think it's good enough for DL ? Thanks in advance. 👍😊
You’re welcome! And yes it’s definitely good enough to get started, I use an M1 MacBook Pro almost everyday
how was your experience after this? Was Machine learning projects going ok in MacOS?
You're awesome! Thank you so much
Thanks!
Really nice video! Have you compared the performance on M1 and Non-M1 MacBook when doing machine learning? M1 MacBook runs how fast?
Thank you Ben! I have compared the M1, M1 Pro and M1 Max machine learning benchmarks, see here:
ua-cam.com/video/JWYsWhR3Pxg/v-deo.html
Hi, i followed exactly how you did, i got this message: 500: Internal Server Error, what to do? 😢
Before Tutorial: No Computer
After Tutorial: Made Self-Driving Car
Hi Daniel, Thanks for wonderful video! Can I ask if you have any video on installing AutoKeras onMacbook M1 ? :)
Not yet, but you should be able to install it into a TensorFlow 2.8+ environment with "pip3 install autokeras"
Thank you, Daniel.
You’re welcome Anfas!
@@mrdbourke FYI. During the process, I had an issue with numpy import error. I had to fix it by installing the default version. Thank you once again!
life changing!
Really nice video.
thanks thanks thanks ! amazing ! just one question. How do I into to env in different way? I can not use " Conda activate env", only Conda activate Users/.../env
You have to make sure you’re in the right folder (where your env is). E.g. if your env is in “TensorFlow-test” folder you need to be in that folder to use “conda activate ./env”
I did everything and everything goes well. However, while I start training, the kernel becomes dead. Any solution that?
Hi Daniel,
Thank you for this video. I have one question How to install other package when needed.
Can you please tell me.
Very useful tutorial. Do you have any idea of how I can install mediapipe in the ./env?
Thanks a million Daniel. Though, I have one question to ask,
Is there any reason that you downloaded Miniforge at the Home directory instead of the root directory?
when I've compared the anaconda location with my other intel devices (where I used brew install) they just downloaded Under the root/opt directory and there seems to be a little difference in making new environments with the following common command
"conda create --name env_name python==version"
Thank you Daniel!
I have NO idea how any of this works and first time I've opened the terminal - and still don't understand any of it, but it works.
How can I get the environment to work on my Spyder installation? In the preferences I can change the environment to "/Users/my_user/tensorflow-test/env/bin/python" and it works with python, but says I do not have TensorFlow installed.
Sorry, but I've only ever worked with graphic interfaces so I have no idea what I'm doing here - but the tutorial was awesome 😊
you may have to make sure Spyder knows that you’re using “conda” as the environment - if you can activate an environment using “conda” in Spyder I’d do that. Perhaps search “use conda environment with Spyder” and then activate the environment we created in this video
Great, thanks! 😊
Could you show us how to clean your HOME directory? After doing all this plus some vim config and stuff, the HOME directory gets messy as fuu
its working pretty good but i have question i have exit the terminal after all this and then when i open the terminal and type jupyter notebook it said invalid command could you please help me with this ?
Thanks Ginger brother !! ❤️️❤️️❤️️❤️️
You're welcome Pierre!
Im getting this error:
could not find a version that satisfies the requiremnt tensorflow-macos (from versions:none)
thanks a lottt!! literally one of the best vids out there (yet need more training to beat Indian youtubers jk) Saved a lot of time and energy, thanks