Remove scale constraint for SUM int8

PiperOrigin-RevId: 754059965

Author: marialyu

ai_edge_quantizer/algorithms/uniform_quantize/common_quantize.py

@@ -416,12 +416,21 @@ def materialize_sum(
     tensor_name_to_qsv: dict[str, Any],
 ) -> list[qtyping.TensorTransformationParams]:
   """Materialize tensors in tfl.sum."""
+  # For 8 bits the reference kernel calls a function without input/output
+  # constraints. For all others it calls a function that enforces input/output
+  # scale/zero point checks. See:
+  # https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/kernels/reduce.cc#L909
+  activation_config = op_info.op_quant_config.activation_tensor_config
+  if activation_config is not None and activation_config.num_bits == 8:
+    constraint = _OpQuantConstraint.NO_CONSTRAIN
+  else:
+    constraint = _OpQuantConstraint.SAME_AS_INPUT_SCALE
   return common_utils.materialize_standard_op(
       op_info,
       graph_info,
       tensor_name_to_qsv,
       get_tensor_quant_params_fn,
-      constraint=_OpQuantConstraint.SAME_AS_INPUT_SCALE,
+      constraint=constraint,
       inputs_to_ignore=[1],  # Axis index does not need to be quantized.
   )
 

ai_edge_quantizer/algorithms/uniform_quantize/op_architecture_tests/sum_test.py

@@ -102,7 +102,7 @@ def test_materialize_sum_succeeds(
         self._op_test_info,
         common_quantize.materialize_sum,
         get_tensor_quant_params_func,
-        same_input_output_params=True,
+        same_input_output_params=num_bits != 8,  # No constraint for 8 bits.
         inputs_to_ignore=[1],  # Ignore axis tensor.
     )