pnnx pass_ncnn 崩溃

[zuowanbushiwo]

error log | 日志或报错信息 | ログ

pnnxparam = dcrn_stft.pnnx.param
pnnxbin = dcrn_stft.pnnx.bin
pnnxpy = dcrn_stft_pnnx.py
pnnxonnx = dcrn_stft.pnnx.onnx
ncnnparam = dcrn_stft.ncnn.param
ncnnbin = dcrn_stft.ncnn.bin
ncnnpy = dcrn_stft_ncnn.py
fp16 = 1
optlevel = 2
device = cpu
inputshape = [1,257]f32,[1,257]f32,[2,1,128]f32,[1,9728]f32,[1,19456]f32
inputshape2 =
customop =
moduleop =
############# pass_level0
inline module = CausalConvBlock
inline module = CausalTransConvBlock
inline module = CausalTransConvBlockEnd
inline module = CausalConvBlock
inline module = CausalTransConvBlock
inline module = CausalTransConvBlockEnd

----------------

############# pass_level1
no attribute value
swapped detected !
no attribute value
############# pass_level2
############# pass_level3
############# pass_level4
############# pass_level5
make_slice_expression input 3
make_slice_expression input 10
make_slice_expression input 3
make_slice_expression input 3
make_slice_expression input 3
make_slice_expression input 3
make_slice_expression input 3
make_slice_expression input 4
make_slice_expression input 4
make_slice_expression input 4
make_slice_expression input 4
make_slice_expression input 4
make_slice_expression input 4
############# pass_ncnn
Segmentation fault (core dumped)

model | 模型 | モデル

1. original model
   Pytorch 模型

how to reproduce | 复现步骤 | 再現方法

1. os.system("~/pnnx-20221116-ubuntu/pnnx dcrn_stft.pt inputshape=[1,257],[1,257],[2,1,128],[1,9728],[1,19456]")
2. 能生成pnnx 相关的文件，使用dcrn_stft_pnnx.py 推理也能得到正确的结果。
3. ncnn 转换失败，ncnn相关的文件没有。
   @nihui 大佬帮看看一下 这个模型可以用ncnn 推理吗

[nihui]

针对onnx模型转换的各种问题，推荐使用最新的pnnx工具转换到ncnn
In view of various problems in onnx model conversion, it is recommended to use the latest pnnx tool to convert your model to ncnn

pip install pnnx
pnnx model.onnx inputshape=[1,3,224,224]

详细参考文档
Detailed reference documentation
https://github.com/pnnx/pnnx
https://github.com/Tencent/ncnn/wiki/use-ncnn-with-pytorch-or-onnx#how-to-use-pnnx

[nihui]

#5779