First step to add forward indexes.

Author: sdpython

tests/test_optimizers.py

@@ -912,7 +912,6 @@ def test_duplicated_duplicated_attributes(self):
                                                 op_type="ReduceSum", remaining_op_num=2)
 
     def _check_initializer_num(self, graph_proto, num):
-        print(len(graph_proto.initializer))
         return num == len(graph_proto.initializer)
 
     def test_duplicated_duplicated_constant(self):

tf2onnx/constants.py

@@ -15,7 +15,7 @@
 MICROSOFT_DOMAIN = "com.microsoft"
 
 # Default opset version for onnx domain
-PREFERRED_OPSET = 8
+PREFERRED_OPSET = 11
 
 # Default opset for custom ops
 TENSORFLOW_OPSET = helper.make_opsetid("ai.onnx.converters.tensorflow", 1)

tf2onnx/graph.py

@@ -418,6 +418,7 @@ def __init__(self, nodes, output_shapes=None, dtypes=None, target=None, opset=No
         self._nodes = []
         self._nodes_by_name = {}
         self._output_to_node_name = {}
+        self._input_to_node_name = {}
         self.shapes = {}
         self.graph_name = graph_name or "tf2onnx"
         self._is_subgraph = is_subgraph
@@ -442,7 +443,7 @@ def __init__(self, nodes, output_shapes=None, dtypes=None, target=None, opset=No
 
         ops = [Node(node, self) for node in nodes]
         self.reset_nodes(ops)
-
+        
         if not is_subgraph:
             # add identity node after each output, in case it is renamed during conversion.
             for o in self.outputs:
@@ -569,6 +570,11 @@ def make_node(self, op_type, inputs, attr=None, output_count=1, outputs=None, sk
 
         onnx_node = helper.make_node(op_type, inputs, outputs, name=name, domain=domain, **raw_attr)
 
+        for name in onnx_node.input:
+            if name not in self._input_to_node_name:
+                self._input_to_node_name[name] = set()
+            self._input_to_node_name[name].add(onnx_node.name)
+
         if op_type in ["If", "Loop", "Scan"]:
             # we force the op containing inner graphs not skipped during conversion.
             skip_conversion = False
@@ -606,6 +612,10 @@ def append_node(self, node):
             self._output_to_node_name[name] = node.name
             self.set_dtype(name, output_dtypes[i])
             self.set_shape(name, output_shapes[i])
+        for name in node.input:
+            if name not in self._input_to_node_name:
+                self._input_to_node_name[name] = set()
+            self._input_to_node_name[name].add(node.name)
 
     def remove_node(self, node_name):
         """Remove node in current graph."""
@@ -626,6 +636,13 @@ def remove_node(self, node_name):
             if op_output in self._dtypes:
                 del self._dtypes[op_output]
 
+        for op_input in node.input:
+            if op_input not in self._input_to_node_name:
+                raise RuntimeError(
+                    "Input %r of node %r not found." % (op_input, node.name))
+            if node.name in self._input_to_node_name[op_input]:
+                self._input_to_node_name[op_input].remove(node.name)
+
         self._nodes.remove(node)
         node.graph = None
 
@@ -649,16 +666,26 @@ def reset_nodes(self, ops):
         self.contained_graphs = remained_sub_graphs
         self._nodes_by_name = {op.name: op for op in ops}
         self._output_to_node_name = {}
+        self._input_to_node_name = {}
         for op in ops:
             for op_output in op.output:
                 self._output_to_node_name[op_output] = op.name
+            for op_input in op.input:
+                if op_input not in self._input_to_node_name:
+                    self._input_to_node_name[op_input] = set()
+                self._input_to_node_name[op_input].add(op.name)
 
         for n in self._order_sensitive_inputs:
             if n not in ops:
                 self._order_sensitive_inputs.remove(n)
         for o in self.outputs:
             if o not in self._output_to_node_name:
-                raise ValueError("graph output " + o + " not exist")
+                raise ValueError("graph output %r not exist" % o)
+        for i in self.inputs:
+            if i.name.startswith('Placeholder'):
+                continue
+            if i.name not in self._input_to_node_name:
+                raise ValueError("graph input %r not exist in graph." % i.name)
 
         self._dtypes = remained_dtypes
         self._output_shapes = remained_shapes
@@ -775,6 +802,14 @@ def get_node_by_output_in_current_graph(self, output):
             ret = self._nodes_by_name.get(name)
         return ret
 
+    def get_node_by_input_in_current_graph(self, input):
+        """Get nodes by node input id."""
+        names = self._output_to_node_name.get(input)
+        ret = None
+        if name:
+            ret = [self._nodes_by_name.get(name) for name in names]
+        return ret
+
     def get_node_by_name(self, name):
         """Get node by name."""
         ret = self._nodes_by_name.get(name)
@@ -785,6 +820,10 @@ def set_node_by_name(self, node):
         self._nodes_by_name[node.name] = node
         for op_output in node.output:
             self._output_to_node_name[op_output] = node.name
+        for name in node.input:
+            if name not in self._input_to_node_name:
+                self._input_to_node_name[name] = set()
+            self._input_to_node_name[name].add(node.name)
 
     def change_node_name(self, node, new_name):
         """Remove node in current graph."""
@@ -1210,35 +1249,54 @@ def find_output_consumers(self, output_name):
                     nodes.extend(g.find_output_consumers(output_name))
         return nodes
 
-    @staticmethod
-    def replace_all_inputs(ops, old_input, new_input):
+    def replace_all_inputs(self, ops, old_input, new_input):
         """Replace all inputs pointing to old_input with new_input."""
         if old_input == new_input:
             return
+        if new_input not in self._input_to_node_name:
+            self._input_to_node_name[new_input] = set()
+        
+        to_ops = self._input_to_node_name.get(old_input, None)
+        if to_ops is None:
+            # This means old_input is a final output.
+            to_ops = set()
 
         for node in ops:
             if old_input in node.input and new_input in node.output:
                 raise RuntimeError("creating a circle in the graph is not allowed: " + node.name)
+            self._input_to_node_name[new_input].add(node.name)
 
             for i, input_name in enumerate(node.input):
                 if input_name == old_input:
                     node.input[i] = new_input
+                    if node.name not in to_ops:
+                        raise RuntimeError(
+                            "Unable to replace %r by %r. Node %r is not using input %r." % (
+                                old_input, new_input, node.name, old_input))
+                
 
             # modify references in sub graphs
             body_graphs = node.get_body_graphs()
             if body_graphs:
                 for g in body_graphs.values():
                     g.replace_all_inputs(g.get_nodes(), old_input, new_input)
 
-    @staticmethod
-    def replace_input(node, old_input, new_input):
+    def replace_input(self, node, old_input, new_input):
         """Replace node."""
         assert isinstance(node, Node) and isinstance(old_input, six.text_type) and isinstance(new_input, six.text_type)
         is_replaced = False
         for i, input_name in enumerate(node.input):
             if input_name == old_input:
                 node.input[i] = new_input
                 is_replaced = True
+        
+        to_ops = self._input_to_node_name.get(old_input, None)
+        if to_ops is not None:
+            to_ops.remove(node.name)
+        if new_input not in self._input_to_node_name:
+            self._input_to_node_name[new_input] = set()
+        self._input_to_node_name[new_input].add(node.name)
+
         return is_replaced
 
     def _extract_sub_graph_nodes(self, dest_node, input_checker=None):

tf2onnx/onnx_opset/controlflow.py

@@ -524,7 +524,7 @@ def version_7(cls, ctx, node, **kwargs):
         maximum_iterations_name = node.input[1]
         maximum_iterations = node.inputs[1].get_tensor_value()
         if maximum_iterations == -1:
-            maximum_iterations = sys.maxsize
+            maximum_iterations = np.iinfo(dtype_loop).max
         consumers = ctx.find_output_consumers(maximum_iterations_name)
         external_consumers = [c for c in consumers if c != node and c.type != 'TensorListReserve']
         if len(external_consumers) == 0: