Merge pull request #158 from fastmachinelearning/feature/remove_forked_identity

Remove identity nodes with output forking

Author: maltanar

src/qonnx/transformation/remove.py

@@ -113,20 +113,20 @@ def apply(self, model):
         graph_modified = False
         for node in graph.node:
             node_ind += 1
-            if node.op_type in ["Add", "Sub"] and not model.is_fork_node(node) and not model.is_join_node(node):
+            if node.op_type in ["Add", "Sub"]:
                 A = model.get_initializer(node.input[1])
                 if A is not None and np.isclose(A, np.zeros_like(A), atol=self.atol).all():
                     remove_node_and_rewire(model, node)
                     graph_modified = True
                     break
 
-            elif node.op_type in ["Mul", "Div"] and not model.is_fork_node(node) and not model.is_join_node(node):
+            elif node.op_type in ["Mul", "Div"]:
                 A = model.get_initializer(node.input[1])
                 if A is not None and np.isclose(A, np.ones_like(A), atol=self.atol).all():
                     remove_node_and_rewire(model, node)
                     graph_modified = True
                     break
-            elif node.op_type == "Pad" and not model.is_fork_node(node) and not model.is_join_node(node):
+            elif node.op_type == "Pad":
                 pads = get_by_name(node.attribute, "pads")
                 if pads is not None:
                     # older versions of Pad op specify pads as attribute

tests/transformation/test_remove_identity_ops.py

@@ -34,13 +34,14 @@
 import qonnx.core.onnx_exec as oxe
 from qonnx.core.datatype import DataType
 from qonnx.core.modelwrapper import ModelWrapper
+from qonnx.transformation.general import SortGraph
 from qonnx.transformation.infer_datatypes import InferDataTypes
 from qonnx.transformation.infer_shapes import InferShapes
 from qonnx.transformation.remove import RemoveIdentityOps
 from qonnx.util.basic import gen_finn_dt_tensor, qonnx_make_model
 
 
-def insert_identity_op(model, op, as_first_node, approx):
+def insert_identity_op(model, op, as_first_node, approx, fork_after_id):
     kwargs = {}
     inp_ndims = 4 if as_first_node else 2
     if approx:
@@ -71,12 +72,24 @@ def insert_identity_op(model, op, as_first_node, approx):
         model.set_initializer("value", val)
         inplist = ["inp" if as_first_node else "div_out", "value"]
     identity_node = helper.make_node(op, inplist, ["ident_out"], **kwargs)
+    old_2nd_node = graph.node[1]
+    old_last_node = graph.node[-1]
+    graph.node.append(identity_node)
+    if fork_after_id:
+        graph.node.append(helper.make_node("Mul", ["ident_out", "mul2"], ["mulbranch0_out"]))
+        model.set_initializer("mul2", np.asarray([2.0], dtype=np.float32))
+        graph.node.append(helper.make_node("Mul", ["ident_out", "mul3"], ["mulbranch1_out"]))
+        model.set_initializer("mul3", np.asarray([3.0], dtype=np.float32))
+        graph.node.append(helper.make_node("Add", ["mulbranch0_out", "mulbranch1_out"], ["idfork_out"]))
+        subgraph_out = "idfork_out"
+    else:
+        subgraph_out = "ident_out"
+
     if as_first_node:
-        graph.node.insert(0, identity_node)
-        graph.node[1].input[0] = "ident_out"
+        old_2nd_node.input[0] = subgraph_out
     else:
-        graph.node.insert(3, identity_node)
-        graph.node[-1].input[0] = "ident_out"
+        old_last_node.input[0] = subgraph_out
+    model = model.transform(SortGraph())
 
     return model
 
@@ -86,7 +99,10 @@ def insert_identity_op(model, op, as_first_node, approx):
 @pytest.mark.parametrize("approx", [False, True])
 @pytest.mark.parametrize("as_first_node", [False, True])
 @pytest.mark.parametrize("fork_before_id", [False, True])
-def test_remove_identity_ops(op, as_first_node, approx, fork_before_id):
+@pytest.mark.parametrize("fork_after_id", [False, True])
+def test_remove_identity_ops(op, as_first_node, approx, fork_before_id, fork_after_id):
+    if approx and not (op in ["Add", "Sub", "Mul", "Div"]):
+        pytest.skip(f"approx=True not relevant for {op}")
     # set up onnx model
     inp = helper.make_tensor_value_info("inp", TensorProto.FLOAT, [1, 4, 1, 1])
     mul = helper.make_tensor_value_info("mul", TensorProto.FLOAT, [])
@@ -119,7 +135,8 @@ def test_remove_identity_ops(op, as_first_node, approx, fork_before_id):
     model.set_initializer("shape", shape_values)
     model.set_initializer("div", div_values)
     model.set_initializer("matmul", matmul_values)
-    insert_identity_op(model, op, as_first_node, approx)
+    insert_identity_op(model, op, as_first_node, approx, fork_after_id)
+    model = model.transform(InferShapes())
     model = model.transform(InferShapes())
     model = model.transform(InferDataTypes())
     idict = {"inp": inp_values}