Method and device for converting MLIR file into PyTorch file, equipment and medium

The invention relates to the technical field of deep learning compiling, and provides a method and device for converting an MLIR file into a PyTorch file, equipment and a medium. The method comprises the following steps: acquiring an MMIR file, and extracting data information and operation information in the MMIR file; the original identifier of the data information is reserved, and the data type of the data information is converted into the Tensor data type in the PyTorch file; constructing a dependency relationship between the operations according to a sequential relationship between the operations; according to the conversion relation between the PyTorch file and the MMIR file, the operation type in the MMIR file is converted into the operation type in the PyTorch file; and forming a PyTorch file according to the converted data type, the dependency relationship between the operations and the converted operation type. According to the method, the MMIR file can be equivalently converted into the PyTorch file