Closed shokoban closed 1 week ago
Thanks for trying out TensorFlow-DirectML-Plugin! Can you confirm that you are running the command pip install tensorflow-cpu
and not pip install tensorflow
. Unfortunately our plugin only supports the cpu version of TensorFlow at this time.
Yes, I was running the command "pip install tensorflow-cpu" and not "pip install tensorflow". I have understand that TensorFlow-DirectML-Plugin only supports the cpu version of TensorFlow. Because I should have found it written in the manual in step4. (Out of curiosity, I have tried "pip install tensorflow", but as you say, I have confirmed that it does not work.)
Just an update: this issue is on the product team subject-matter experts' radar. They have all the info from this thread. So I'll go ahead and close this issue until such time as they have the bandwidth to address it.
-Steve
[Enter feedback here]
page
checked date
comments
in STEP 4: Install base TensorFlow
The command in the current documentation downloads sensorflow==2.10.0. However, when I install the directml plugin into this version, it does not run wll. It will give error messages such as "relu function not implemented. Therefore, it was necessary to specify the version as follows.
cofirmation environment
PC1
PC2
use trained model
model_vgg16 = tf.keras.applications.vgg16.VGG16(include_top=False, pooling='avg')
model = keras.Sequential() model.add(model_vgg16) model.add(Dense(128, activation="relu")) model.add(Dense(8, activation="softmax"))
model.compile(loss='categorical_crossentropy', optimizer=keras.optimizers.SGD(lr=1e-3, decay=1e-6, momentum=0.9, nesterov=True), metrics=['accuracy'])