chocolatedesue / vits_web_demo Goto Github PK

    seq = torch.randint(low=0, high=phone_num, size=(1, 10), dtype=torch.long)
    seq_len = torch.IntTensor([seq.size(1)]).long()
    print(f"{seq.size()}")
    print(f"{seq}")
    print("--------------------")
    print(f"{seq_len.size()}")
    print(f"{seq_len}")


    # noise(可用于控制感情等变化程度) lenth(可用于控制整体语速) noisew(控制音素发音长度变化程度)
    # 参考 https://github.com/gbxh/genshinTTS
    scales = torch.FloatTensor([0.667, 1.0, 0.8])
    # make triton dynamic shape happy
    scales = scales.unsqueeze(0)
    sid = torch.IntTensor([0]).long()
    # sid=torch.LongTensor([0])
    print(f"{scales.size()}")
    print(f"{sid.size()}")
    tmp=to_numpy(sid)
    print(f"{type(sid)}, tmp={type(tmp)}")



    dummy_input = (seq, seq_len, scales, sid)
    torch.onnx.export(model=net_g,
                      args=dummy_input,
                      f=args.onnx_model,
                      input_names=['input', 'input_lengths', 'scales', 'sid'],
                      output_names=['output'],
                      dynamic_axes={
                          'input': {
                              0: 'batch',
                              1: 'phonemes'
                          },
                          'input_lengths': {
                              0: 'phonemes_len'
                          },
                          'scales': {
                              0: 'batch'
                          },
                          'sid': {
                              0: 'batch'
                          },
                          'output': {
                              0: 'batch',
                              1: 'audio',
                              2: 'audio_length'
                          }
                      },
                      opset_version=11,
                      verbose=False)
    # Test case
    seq = torch.randint(low=0, high=phone_num, size=(1, 15), dtype=torch.long)
    seq_len = torch.IntTensor([seq.size(1)]).long()

    # Verify onnx precision
    print(f"In Multi speaker {num_speakers}")
    torch_output = net_g(seq, seq_len, scales, sid)
    providers = [args.providers]
    ort_sess = ort.InferenceSession(args.onnx_model, providers=providers)
    ort_inputs = {
            'input': to_numpy(seq),
            'input_lengths': to_numpy(seq_len),
            'scales': to_numpy(scales),
            'sid': to_numpy(sid),
    }

2023-06-05 09:18:47.766977537 [E:onnxruntime:, sequential_executor.cc:339 Execute] Non-zero status code returned while running Expand node. Name:'Expand_2897' Status Message: invalid expand shape
Traceback (most recent call last):
  File "export_onnx.py", line 203, in <module>
    main()
  File "export_onnx.py", line 195, in main
    audio = np.squeeze(ort_sess.run(None, ort_inputs))
  File "/usr/local/lib/python3.6/dist-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 188, in run
    return self._sess.run(output_names, input_feed, run_options)
onnxruntime.capi.onnxruntime_pybind11_state.InvalidArgument: [ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Non-zero status code returned while running Expand node. Name:'Expand_2897' Status Message: invalid expand shape