1.onnx参数说明
- 其中的input_size_list必须是静态参数
2.模型参数查看脚本
示例输出：
=== 输入节点 === Name: images Shape: [1, 3, 640, 640] === 输出节点 === Name: output Shape: [1, 255, 80, 80] Name: 283 Shape: [1, 255, 40, 40] Name: 285 Shape: [1, 255, 20, 20]import onnx # 加载模型 model = onnx.load('yolov5s_relu.onnx') # 查看所有输入 print("=== 输入节点 ===") for input in model.graph.input: print(f"Name: {input.name}") # 获取 shape shape = [dim.dim_value if dim.dim_value else dim.dim_param for dim in input.type.tensor_type.shape.dim] print(f"Shape: {shape}") # print(f"Dtype: {input.type.tensor_type.elem_type}") # 1=FLOAT, 7=INT64 # 查看所有输出 print("\n=== 输出节点 ===") for output in model.graph.output: print(f"Name: {output.name}") shape = [dim.dim_value if dim.dim_value else dim.dim_param for dim in output.type.tensor_type.shape.dim] print(f"Shape: {shape}") # 查看所有中间节点（可选） # print("\n=== 所有节点 ===") # for node in model.graph.node: # print(f"{node.op_type}: {node.name}")3.模型转换脚本
from rknn.api import RKNN rknn = RKNN(verbose=True) rknn.config( target_platform='rk3568', ) batch_size=1 sequence_length=512 past_sequence_length=512 ret = rknn.load_onnx(model='yolov5s_relu.onnx') if ret != 0: print('Load model failed!') exit(ret) print('done') ret = rknn.build(do_quantization=False) if ret != 0: print('Load model failed!') exit(ret) print('done') ret = rknn.export_rknn(export_path='./rknn.rknn') if ret != 0: print('Load model failed!') exit(ret) print('done')
