Fix PReLU conversion from CoreML (#425)

Signed-off-by: Michael Dudalev <michael@dudalev.com>

Co-authored-by: Wenbing Li <10278425+wenbingl@users.noreply.github.com>

onnxmltools/convert/coreml/operator_converters/neural_network/Activation.py

@@ -4,8 +4,10 @@
 # license information.
 # --------------------------------------------------------------------------
 
+import numpy as np
+
 from ....common._apply_operation import apply_elu, apply_hard_sigmoid, apply_leaky_relu, apply_prelu, apply_relu, \
-    apply_sigmoid, apply_tanh, apply_affine, apply_parametric_softplus, apply_scaled_tanh
+    apply_sigmoid, apply_tanh, apply_affine, apply_parametric_softplus, apply_scaled_tanh, apply_thresholded_relu
 from ....common._registration import register_converter
 
 
@@ -24,7 +26,7 @@ def convert_activation(scope, operator, container):
     elif activation_type == 'ReLU':
         apply_relu(scope, inputs, outputs, container, operator_name=attrs['name'])
     elif activation_type == 'PReLU':
-        apply_prelu(scope, inputs, outputs, container, operator_name=attrs['name'], slope=[params.PReLU.alpha])
+        apply_prelu(scope, inputs[0], outputs, container, operator_name=attrs['name'], slope=np.asarray([params.PReLU.alpha.floatValue]))
     elif activation_type == 'ELU':
         apply_elu(scope, inputs, outputs, container, operator_name=attrs['name'], alpha=params.ELU.alpha)
     elif activation_type == 'tanh':