Hi everybody,
I was excited to check out Tensorflow for M1, as advertised. I've got the latest version to run by following
https://developer.apple.com/metal/tensorflow-plugin/
Then, I did a little performance check by running the code from
https://www.tensorflow.org/tutorials/quickstart/beginner
To my surprise, with tensorflow-metal installed, an epoch takes 7-8 seconds to complete in average. Without tensorflow-metal installed, it just takes 1 second.
Here the output I'm getting:
Metal device set to: Apple M1
systemMemory: 16.00 GB maxCacheSize: 5.33 GB
2021-11-12 19:55:13.379972: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:305] Could not identify NUMA node of platform GPU ID 0, defaulting to 0. Your kernel may not have been built with NUMA support.
2021-11-12 19:55:13.380077: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:271] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 0 MB memory) -> physical PluggableDevice (device: 0, name: METAL, pci bus id: )
[[0.05593071 0.11313408 0.0657519 0.04874061 0.11589434 0.1239315
0.09336308 0.1969078 0.13757557 0.04877035]]
2021-11-12 19:55:13.632319: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:185] None of the MLIR Optimization Passes are enabled (registered 2)
2021-11-12 19:55:13.632493: W tensorflow/core/platform/profile_utils/cpu_utils.cc:128] Failed to get CPU frequency: 0 Hz
Epoch 1/5
2021-11-12 19:55:13.721079: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:112] Plugin optimizer for device_type GPU is enabled.
1875/1875 [==============================] - 7s 4ms/step - loss: 0.2888 - accuracy: 0.9160
Epoch 2/5
1875/1875 [==============================] - 7s 4ms/step - loss: 0.1419 - accuracy: 0.9577
Epoch 3/5
1875/1875 [==============================] - 7s 4ms/step - loss: 0.1057 - accuracy: 0.9679
Epoch 4/5
1875/1875 [==============================] - 7s 4ms/step - loss: 0.0877 - accuracy: 0.9733
Epoch 5/5
1875/1875 [==============================] - 7s 4ms/step - loss: 0.0729 - accuracy: 0.9771
2021-11-12 19:55:47.920935: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:112] Plugin optimizer for device_type GPU is enabled.
Best from Berlin, Michael
