NXP backend: Condition fused bias quantization pass to QAT flow only

Author: StrycekSimon

backends/nxp/quantizer/utils.py

@@ -219,8 +219,9 @@ def calibrate_and_quantize(
 
     m = convert_pt2e(m)
 
-    m = QuantizeFusedConvBnBiasAtenPass(default_zero_bias=False, symmetric_quant=True)(
-        m
-    ).graph_module
+    if is_qat:
+        m = QuantizeFusedConvBnBiasAtenPass(
+            default_zero_bias=False, symmetric_quant=True
+        )(m).graph_module
 
     return m