Merge pull request #866 from mmrahorovic/fix/minimize_bitwidth_thresh…

…olds Fix to threshold's width optimization
Xilinx · Aug 4, 2023 · fd72d48 · fd72d48
2 parents 99f61fc + f52871d
commit fd72d48
Show file tree

Hide file tree

Showing 3 changed files with 10 additions and 2 deletions.
diff --git a/src/finn/custom_op/fpgadataflow/matrixvectoractivation.py b/src/finn/custom_op/fpgadataflow/matrixvectoractivation.py
@@ -659,7 +659,6 @@ def minimize_accumulator_width(self, model):
             # for no-activation nodes, output dt = acc dt
             self.set_nodeattr("outputDataType", adt.name)
         self.set_nodeattr("accDataType", adt.name)
-
         return DataType[self.get_nodeattr("accDataType")]
 
     def minimize_weight_bit_width(self, model):

diff --git a/src/finn/custom_op/fpgadataflow/thresholding_batch.py b/src/finn/custom_op/fpgadataflow/thresholding_batch.py
@@ -211,6 +211,8 @@ def minimize_accumulator_width(self, model):
             threshold_tensor
         ).all(), "Thresholds can't be expressed with type %s" % str(tdt)
         self.set_nodeattr("weightDataType", tdt.name)
+        # Update QONNX DataType of tensor for consistency
+        model.set_tensor_datatype(self.onnx_node.input[1], tdt)
         return DataType[self.get_nodeattr("weightDataType")]
 
     def get_instream_width(self, ind=0):

diff --git a/src/finn/transformation/fpgadataflow/minimize_accumulator_width.py b/src/finn/transformation/fpgadataflow/minimize_accumulator_width.py
@@ -28,6 +28,7 @@
 
 from qonnx.custom_op.registry import getCustomOp
 from qonnx.transformation.base import Transformation
+from qonnx.transformation.infer_datatypes import InferDataTypes
 
 from finn.util.fpgadataflow import is_fpgadataflow_node
 
@@ -41,9 +42,15 @@ def __init__(self):
         super().__init__()
 
     def apply(self, model):
-        for node in model.graph.node:
+        for node_id in range(len(model.graph.node)):
+            # Since InferDataTypes potentially changes node attributes in each loop iterations,
+            # the for-loop cannot loop over a list of a snapshot of the graph's node protos
+            node = model.graph.node[node_id]
             if is_fpgadataflow_node(node) is True:
                 inst = getCustomOp(node)
                 if hasattr(inst, "minimize_accumulator_width"):
                     inst.minimize_accumulator_width(model)
+                    # Since this transformation is applied iteratively, we have to ensure that
+                    # we propagate the new datatype to other layers
+                    model = model.transform(InferDataTypes())
         return (model, False)