`import tensorflow as tf tf.config.list_physical_devices() [PhysicalDevice(name='/physical_device:CPU:0', device_type='CPU'), PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]` As soon as I try to run a keras based model it dies with: 2021-11-08 19:11:56.350233: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: SSE4.2 AVX AVX2 FMA To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags. 2021-11-08 19:11:56.350804: 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. 2021-11-08 19:11:56.351033: 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: ) 2021-11-08 19:11:56.512351: I tensorflow/core/profiler/lib/profiler_session.cc:131] Profiler session initializing. 2021-11-08 19:11:56.512369: I tensorflow/core/profiler/lib/profiler_session.cc:146] Profiler session started. 2021-11-08 19:11:56.512818: I tensorflow/core/profiler/lib/profiler_session.cc:164] Profiler session tear down. 2021-11-08 19:11:57.362096: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:185] None of the MLIR Optimization Passes are enabled (registered 2)

Metal device set to: AMD Radeon Pro 5600M 2022-01-13 17:02:36.447465: I tensorflow/core/platform/cpu_feature_guard.cc:151] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 FMA To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags. 2022-01-13 17:02:36.448221: 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. 2022-01-13 17:02:36.448581: 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: ) Prior to running my model: print("Num GPUs Available: ", len(tf.config.experimental.list_physical_devices('GPU'))) Num GPUs Available: 1 I was excited to see an tensorflow-apple version 2.7 but STILL does not work.

I'll add that like others, python 3.8.X is required. Then run: SYSTEM_VERSION_COMPAT=0 python -m pip install tensorflow-macos (this will give you tensorflow-macos version 2.7 - don't get excited yet) SYSTEM_VERSION_COMPAT=0 python -m pip install tensorflow-metal (this will also succeed) Install your other stuff as normal with conda (pandas, scikit-learn, jupyterlab, etc) Then run your code. You can run: print("Num GPUs Available: ", len(tf.config.experimental.list_physical_devices('GPU'))) Num GPUs Available: 1 So far so good right?!?! Looking Good! Now run your model... First line: Metal device set to: AMD Radeon Pro 5600M keeps looking better and better !!! and then the kernel dies... 2022-01-13 17:02:36.447465: I tensorflow/core/platform/cpu_feature_guard.cc:151] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 FMA To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags. 2022-01-13 17:02:36.448221: 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. 2022-01-13 17:02:36.448581: 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: <undefined>) Will tensorflow-macos ever work for Intel ?

I tried running my script outside of jupyter lab and the error is as follows: 2022-03-29 13:26:37.306914: 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. 2022-03-29 13:26:37.307420: 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: ) Model: "sequential" Layer (type) Output Shape Param # masking (Masking) (None, 1, 28) 0 layer1 (Bidirectional) (None, 1, 128) 47616 dropout (Dropout) (None, 1, 128) 0 layer2 (Bidirectional) (None, 1, 128) 98816 dropout_1 (Dropout) (None, 1, 128) 0 layer3 (Bidirectional) (None, 128) 98816 Output (Dense) (None, 1) 129 ================================================================= Total params: 245,377 Trainable params: 245,377 Non-trainable params: 0 Epoch 1/3000 2022-03-29 13:26:54.863499: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:113] Plugin optimizer for device_type GPU is enabled. 2022-03-29 13:26:57.201 python[66880:3338950] -[MPSGraph adamUpdateWithLearningRateTensor:beta1Tensor:beta2Tensor:epsilonTensor:beta1PowerTensor:beta2PowerTensor:valuesTensor:momentumTensor:velocityTensor:gradientTensor:name:]: unrecognized selector sent to instance 0x600041707c60 zsh: segmentation fault python test.py I was able to run another tensorflow based script on cmd line so there is something specific to this error that I am getting.

I have a similar issue with kernel crashing with Adam: Model: "sequential" Layer (type) Output Shape Param # masking (Masking) (None, 1, 28) 0 layer1 (Bidirectional) (None, 1, 128) 47616 dropout (Dropout) (None, 1, 128) 0 layer2 (Bidirectional) (None, 1, 128) 98816 dropout_1 (Dropout) (None, 1, 128) 0 layer3 (Bidirectional) (None, 128) 98816 Output (Dense) (None, 1) 129 ================================================================= Total params: 245,377 Trainable params: 245,377 Non-trainable params: 0 Epoch 1/3000 2022-03-29 13:26:54.863499: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:113] Plugin optimizer for device_type GPU is enabled. **2022-03-29 13:26:57.201 python[66880:3338950] -[MPSGraph adamUpdateWithLearningRateTensor:beta1Tensor:beta2Tensor:epsilonTensor:beta1PowerTensor:beta2PowerTensor:valuesTensor:momentumTensor:velocityTensor:gradientTensor:name:]: unrecognized selector sent to instance 0x600041707c60 ** zsh: segmentation fault python test.py

I tried swapping Adam for RMSprop - still get an error - but now it's a floating point error. 2022-03-29 13:35:35.219059: 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. 2022-03-29 13:35:35.219398: 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: ) Model: "sequential" Layer (type) Output Shape Param # masking (Masking) (None, 1, 28) 0 layer1 (Bidirectional) (None, 1, 128) 47616 dropout (Dropout) (None, 1, 128) 0 layer2 (Bidirectional) (None, 1, 128) 98816 dropout_1 (Dropout) (None, 1, 128) 0 layer3 (Bidirectional) (None, 128) 98816 Output (Dense) (None, 1) 129 ================================================================= Total params: 245,377 Trainable params: 245,377 Non-trainable params: 0 Epoch 1/3000 2022-03-29 13:35:51.773828: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:113] Plugin optimizer for device_type GPU is enabled. 2022-03-29 13:35:55.049448: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:113] Plugin optimizer for device_type GPU is enabled. 2022-03-29 13:35:55.409149: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:113] Plugin optimizer for device_type GPU is enabled. 2022-03-29 13:35:58.358459: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:113] Plugin optimizer for device_type GPU is enabled. 2022-03-29 13:35:58.544457: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:113] Plugin optimizer for device_type GPU is enabled. 2022-03-29 13:35:58.809214: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:113] Plugin optimizer for device_type GPU is enabled. 2022-03-29 13:35:58.989445: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:113] Plugin optimizer for device_type GPU is enabled. 2022-03-29 13:36:01.117681: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:113] Plugin optimizer for device_type GPU is enabled. 2022-03-29 13:36:01.338093: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:113] Plugin optimizer for device_type GPU is enabled. zsh: floating point exception python test.py

I have that exact same problem: [MPSGraph adamUpdateWithLearningRateTensor:beta1Tensor:beta2Tensor:epsilonTensor:beta1PowerTensor:beta2PowerTensor:valuesTensor:momentumTensor:velocityTensor:gradientTensor:name:]: unrecognized selector sent to instance 0x60002836b9c0 I tried switching Adam to RMSprop and then I get floating point exception.

tensorboard 2.11.2 pypi_0 pypi tensorboard-data-server 0.6.1 pypi_0 pypi tensorboard-plugin-wit 1.8.1 pypi_0 pypi tensorflow-estimator 2.11.0 pypi_0 pypi tensorflow-io-gcs-filesystem 0.29.0 pypi_0 pypi tensorflow-macos 2.11.0 pypi_0 pypi tensorflow-metal 0.7.0 pypi_0 pypi Running the same python test script at the apple metal page: 2023-01-20 12:52:34.536215: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: SSE4.1 SSE4.2 AVX AVX2 FMA To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags. Downloading data from https://www.cs.toronto.edu/~kriz/cifar-100-python.tar.gz 169001437/169001437 [==============================] - 7s 0us/step 2023-01-20 12:53:02.967585: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: SSE4.1 SSE4.2 AVX AVX2 FMA To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags. Metal device set to: AMD Radeon Pro 5600M systemMemory: 64.00 GB maxCacheSize: 3.99 GB 2023-01-20 12:53:02.968211: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:306] Could not identify NUMA node of platform GPU ID 0, defaulting to 0. Your kernel may not have been built with NUMA support. 2023-01-20 12:53:02.968256: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:272] 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: ) Epoch 1/5 /opt/anaconda3/envs/applemetal/lib/python3.10/site-packages/keras/backend.py:5585: UserWarning: "sparse_categorical_crossentropy received from_logits=True, but the output argument was produced by a Softmax activation and thus does not represent logits. Was this intended? output, from_logits = _get_logits( 2023-01-20 12:53:16.475463: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:114] Plugin optimizer for device_type GPU is enabled. 2023-01-20 12:53:25.908578: W tensorflow/core/framework/op_kernel.cc:1830] OP_REQUIRES failed at xla_ops.cc:418 : NOT_FOUND: could not find registered platform with id: 0x7f9194090a60 2023-01-20 12:53:25.908651: W tensorflow/core/framework/op_kernel.cc:1830] OP_REQUIRES failed at xla_ops.cc:418 : NOT_FOUND: could not find registered platform with id: 0x7f9194090a60 .... Traceback (most recent call last): File "/Users/ray/test.py", line 13, in model.fit(x_train, y_train, epochs=5, batch_size=64) File "/opt/anaconda3/envs/applemetal/lib/python3.10/site-packages/keras/utils/traceback_utils.py", line 70, in error_handler raise e.with_traceback(filtered_tb) from None File "/opt/anaconda3/envs/applemetal/lib/python3.10/site-packages/tensorflow/python/eager/execute.py", line 52, in quick_execute tensors = pywrap_tfe.TFE_Py_Execute(ctx._handle, device_name, op_name, tensorflow.python.framework.errors_impl.NotFoundError: Graph execution error: Detected at node 'StatefulPartitionedCall_212' defined at (most recent call last): File "/Users/ray/test.py", line 13, in model.fit(x_train, y_train, epochs=5, batch_size=64) File "/opt/anaconda3/envs/applemetal/lib/python3.10/site-packages/keras/utils/traceback_utils.py", line 65, in error_handler return fn(*args, **kwargs) File "/opt/anaconda3/envs/applemetal/lib/python3.10/site-packages/keras/engine/training.py", line 1650, in fit tmp_logs = self.train_function(iterator) File "/opt/anaconda3/envs/applemetal/lib/python3.10/site-packages/keras/engine/training.py", line 1249, in train_function .... ....