修复生成onnx格式错误的bug

examples/pytorch/resnet50/export_resnet50_2_onnx.py

@@ -1,6 +1,22 @@
 import torch
 import torchvision.models as models
-model = models.resnet50()
-input_tensor = torch.rand(1,3,224,224)
-model_onnx = torch.jit.trace(model, input_tensor)
-model_onnx.save('resnet50.onnx')
+
+# 加载模型
+model = models.resnet50(pretrained=True)
+model.eval()
+
+# 构造输入
+dummy_input = torch.randn(1, 3, 224, 224)
+
+# 导出 ONNX
+torch.onnx.export(
+    model,
+    dummy_input,
+    "resnet50.onnx",
+    export_params=True,
+    opset_version=11,
+    do_constant_folding=True,
+    input_names=["input"],
+    output_names=["output"],
+    dynamic_axes={"input": {0: "batch_size"}, "output": {0: "batch_size"}}
+)