Add PengChengStarling models to sherpa-onnx (#1835)

scripts/peng-cheng-starling/quantize_models.py

@@ -0,0 +1,22 @@
+#!/usr/bin/env python3
+from onnxruntime.quantization import QuantType, quantize_dynamic
+from pathlib import Path
+
+
+def main():
+    suffix = "epoch-75-avg-11-chunk-16-left-128"
+
+    for m in ["encoder", "joiner"]:
+        if Path(f"{m}-{suffix}.int8.onnx").is_file():
+            continue
+
+        quantize_dynamic(
+            model_input=f"./{m}-{suffix}.onnx",
+            model_output=f"./{m}-{suffix}.int8.onnx",
+            op_types_to_quantize=["MatMul"],
+            weight_type=QuantType.QInt8,
+        )
+
+
+if __name__ == "__main__":
+    main()