使用session = onnxruntime.InferenceSession(args.model,providers=['TIDLCompilationProvider'])
onnxruntime部分 session =onnxruntime.InferenceSession(args.model,providers=['TIDLExecutionProvider'])
还是说必须使用edge-ai重新训练一个模型以得到两个yaml文件
我参考了edgeai-dl-infer的py文件，但是缺少"artifacts_folder": artifacts,"core_number": core_number, 也没有找到readme之类的参考文档，请求回复
"""
Create a TFLite runtime object
Args:
artifacts: TIDL artifacts, can be None if enable_tidl == false
model_path: Path to tflite model file
enable_tidl: Enable TIDL acceleration
core_number: Core Number
"""
delegate_options = {
"tidl_tools_path": "null",
"artifacts_folder": artifacts,
"import": "no",
"core_number": core_number,
}
tidl_delegate = [
_tflitert_interpreter.load_delegate(
"/usr/lib/libtidl_tfl_delegate.so", delegate_options
)
]
self.interpreter = _tflitert_interpreter.Interpreter(
model_path, experimental_delegates=tidl_delegate
)
