Linghao Yu

Results 5 issues of Linghao Yu

您好 我使用您的checkpoint在ICASSP 2021 AEC Challenge blind_test_set clean上验证,没有取得相同的得分 ICASSP 2021 AEC Challenge blind_test_set clean: https://github.com/microsoft/AEC-Challenge/tree/20274d3a85f752d3fa41ed8e8915e73a5e39b7ff/datasets/blind_test_set/clean AECMOS: https://github.com/microsoft/AEC-Challenge/blob/main/AECMOS/AECMOS_local/Run_1663915512_Stage_0.onnx echo score: 3.32 degradation score: 3.10

I try to convert a model called DPRNN with pnnx, it did not report error, but the output shape is inconsistent with input shape. Moreover, if I permute it to...

因为默认是False,是不是应该手动设置一下True

首先非常感谢您的工作,我这里使用onnxsim.simplify导出时会报这个错 `RuntimeError: /project/third_party/onnx-optimizer/onnxoptimizer/passes/eliminate_shape_gather.h:48: runTransform: Assertion 'indices_val < dims.size()' failed.` 我和DeepVQE对比后排查了如下的地方: chunk改切片 转置卷积去掉 groups去掉 空洞卷积去掉 layernorm去掉 均没有排查到原因,还是报上述的错 请问您找到问题在哪了吗

1. SFE模块的unfold可以用如下模块代替,可以减少很多算子 ``` import torch import torch.nn as nn class Unfold(nn.Module): def __init__(self): super().__init__() kernel = torch.eye(3) kernel = kernel.view(3, 1, 1, 3) kernel = nn.Parameter(kernel.repeat(8, 1, 1, 1)) self.conv...