Update run_generation_cpu_woq.py (#1537)

Signed-off-by: Wang, Chang <chang1.wang@intel.com>

examples/huggingface/pytorch/text-generation/quantization/run_generation_cpu_woq.py

@@ -163,7 +163,12 @@
 parser.add_argument("--use_neural_speed", action="store_true")
 # =======================================
 args = parser.parse_args()
-args.scheme = "asym" if args.scheme is None and args.woq_algo == "AutoRound" else "sym" 
+# woq AutoRound algo expects the default scheme is asym, others are sym.
+if args.scheme is None:
+    if args.woq_algo == "AutoRound":
+        args.scheme = "asym"
+    else:
+        args.scheme = "sym"
 
 config = AutoConfig.from_pretrained(
     args.model,