multi_gpu.py

# Define multi-gpu strategy 
mirrored_strategy = tf.distribute.MirroredStrategy()
# Update batch size value
batch_size *= mirrored_strategy.num_replicas_in_sync
# Create strategy scope to perform training
with mirrored_strategy.scope():
    model = [...]
    model.fit(...)

Is the model.fit(...) called within the scope of
with mirrored_strategy.scope():
or outside of it?

Is the model.fit(...) called within the scope of
with mirrored_strategy.scope():
or outside of it?

It is called within the scope.