在使用rknn=1.6.0，load_onnx模型转换时，选用“mmse”的量化方式时，会包下面的错误。使用“normal”的方式是正常的。跪求解答。

[yukaizhou]

D Real output shape: (2, 4, 4, 1792) I Build torch-jit-export complete. I Clean. MMSE Quant Step 0: 96%|████████████████████████████████████████▏ | 316/330 [00:30<00:01, 8.51it/s]D Init sub net for ['Mul_Mul_350_10'] D import clients finished I Mul_Mul_350_100@Conv_Conv_348_13:out0 I Mul_Mul_350_101@Initializer_1170_14:out0 E Can not find preprocess parameters for Initializer_1170_14. W ----------------Warning(9)---------------- MMSE Quant Step 0: 96%|████████████████████████████████████████▎ | 317/330 [00:30<00:01, 10.26it/s] E Catch exception when building RKNN model! E Traceback (most recent call last): E File "rknn/api/rknn_base.py", line 592, in rknn.api.rknn_base.RKNNBase.build E File "rknn/api/rknn_base.py", line 1490, in rknn.api.rknn_base.RKNNBase._quantize2 E File "rknn/base/RKNNlib/RK_nn.py", line 553, in rknn.base.RKNNlib.RK_nn.RKnn.quantize E File "rknn/base/RKNNlib/app/medusa/quantization.py", line 145, in rknn.base.RKNNlib.app.medusa.quantization.Quantization.run E File "rknn/base/RKNNlib/app/medusa/quantization.py", line 63, in rknn.base.RKNNlib.app.medusa.quantization.Quantization._run_quantization E File "rknn/base/RKNNlib/app/medusa/workspace.py", line 430, in rknn.base.RKNNlib.app.medusa.workspace.Workspace.run E File "rknn/base/RKNNlib/app/medusa/workspace.py", line 288, in rknn.base.RKNNlib.app.medusa.workspace.Workspace._mmse_init_sub_net E File "rknn/api/rknn_log.py", line 312, in rknn.api.rknn_log.RKNNLog.e E ValueError: Can not find preprocess parameters for Initializer_1170_14.

[eRaul]

RKNN Toolkit已经更新到1.7.5了，里面有解一些MMSE的bug，可以更新工具后再试下。