optimize TFPark create tf model performance (intel-analytics#2694)

* optimize create tf model

* fix style

* fix style

python/orca/src/bigdl/orca/tfpark/tf_optimizer.py

@@ -113,16 +113,16 @@ def __init__(self, training_helper_layer, criterion, val_methods):
     def _expand_inputs(inputs, tensors_with_value, loss):
         additional_inputs = []
         additional_values = []
-        all_required_inputs = find_placeholders([loss])
-        all_required_inputs_names = [v.name for v in all_required_inputs]
+        inputs = nest.flatten(inputs)
+        names = set([i.name for i in inputs])
+
         if tensors_with_value:
             for t, v in tensors_with_value.items():
-                if t.name in all_required_inputs_names:
-                    additional_inputs.append(t)
-                    additional_values.append(v)
-
-        if not isinstance(inputs, list):
-            inputs = nest.flatten(inputs)
+                if t.name in names:
+                    msg = f"tensor {t} already in inputs, cannot put it in tensor_with_value"
+                    raise ValueError(msg)
+                additional_inputs.append(t)
+                additional_values.append(v)
 
         return inputs, additional_inputs, additional_values
 
@@ -442,7 +442,6 @@ def _get_vars_grads(loss):
     def _get_vars_grads_from_train_op(train_op):
         def predicate(t):
             return t.name.split("/")[-1].startswith("zoo_identity_op_for_grad")
-
         grads = find_tensors([train_op], predicate)
         grad_ops = [grad.op for grad in grads]
         variables = []