fairseq
fairseq copied to clipboard
Inferring the decoder and encoder of a transformer model in a streaming manner
Is it possible to infer the model separately through encoder.onnx and decoder.onnx?