You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Having set up the osld environment following steps below (see screenshot), I was able to train HMMs on my M1 macbook. However, tensorflow did not seem to use GPU at all.
Solution
As suggested by https://developer.apple.com/metal/tensorflow-plugin/ , seems like installing the tensorflow-metal plugin solves the problem. Concretely, after pip install tensorflow-macos==2.9.1, I also ran pip install tensorflow-metal==0.5.0. This works in my case as in, if I evaluate
import tensorflow as tf
tf.config.list_physical_devices('GPU')
I now get [PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')].
The text was updated successfully, but these errors were encountered:
Short description of the problem
Having set up the
osld
environment following steps below (see screenshot), I was able to train HMMs on my M1 macbook. However, tensorflow did not seem to use GPU at all.Solution
As suggested by https://developer.apple.com/metal/tensorflow-plugin/ , seems like installing the
tensorflow-metal
plugin solves the problem. Concretely, afterpip install tensorflow-macos==2.9.1
, I also ranpip install tensorflow-metal==0.5.0
. This works in my case as in, if I evaluateI now get
[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]
.The text was updated successfully, but these errors were encountered: