Update download_model.py

download_model.py

@@ -1,4 +1,5 @@
 from huggingface_hub import hf_hub_download
+from onnxruntime.quantization import quantize_dynamic, QuantType
 import shutil
 import os
 
@@ -25,8 +26,22 @@ for cfg in configs:
     except Exception as e:
         print(f"❌ Failed: {e}")
 
-if model_path:
-    shutil.copy(model_path, "./model/model.onnx")
-    print("✅ ONNX model saved to ./model/model.onnx")
-else:
-    raise RuntimeError("Could not find model.onnx in any repo type")
+if not model_path:
+    raise RuntimeError("Could not find model.onnx in any repo type")
+
+# Save original first
+print("Saving original model...")
+shutil.copy(model_path, "./model/model_original.onnx")
+
+# Quantize to INT8
+print("Quantizing to INT8...")
+quantize_dynamic(
+    model_input="./model/model_original.onnx",
+    model_output="./model/model.onnx",
+    weight_type=QuantType.QUInt8,
+    optimize_model=True
+)
+
+# Remove original to save space
+os.remove("./model/model_original.onnx")
+print("✅ Quantized ONNX model saved to ./model/model.onnx")