我的环境下，读取模型文件时发生了错误

[dlmuzpy]

感谢大佬工作！ 我好像读不出来你的模型
python3命令行下，使用torch 1.6版本

输入：
import torch
model = torch.load('yolov4-p5.pt')

返回如下错误:
File "/home/popzq/.conda/envs/zpy/lib/python3.7/site-packages/torch/serialization.py", line 765, in _legacy_load
result = unpickler.load()
ModuleNotFoundError: No module named 'mc'

[BaofengZan]

一样遇到了，不知道是不是需要最新的pytorch

[pingdow]

一样遇到了

[WongKinYiu]

installation is updated.

[pingdow]

installation is updated.

Thank you. I made it

[wuzhenxin1989]

csrc/cpu/mish_cpu.obj build\temp.win-amd64-3.8\Release\csrc/cuda/mish_cuda.obj build\temp.win-amd64-3.8\Release\csrc/cuda/mish_kernel.obj /OUT:build\lib.win-amd64-3.8\mish_cuda_C.cp38-win_amd64.pyd /IMPLIB:build\temp.win-amd64-3.8\Release\csrc/cpu_C.cp38-win_amd64.lib
正在创建库 build\temp.win-amd64-3.8\Release\csrc/cpu_C.cp38-win_amd64.lib 和对象 build\temp.win-amd64-3.8\Release\csrc/cpu_C.cp38-win_amd64.exp
正在生成代码
已完成代码的生成

model = attempt_load(weights, map_location=device)  # load FP32 model

File "C:\Users\Administrator.SC-201905102106\Desktop\ScaledYOLOv4-yolov4-large\models\experimental.py", line 137, in attempt_load
model.append(torch.load(w, map_location=map_location)['model'].float().fuse().eval()) # load FP32 model
File "D:\Program Files\Anaconda36\envs\yolov52\lib\site-packages\torch\serialization.py", line 593, in load
return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
File "D:\Program Files\Anaconda36\envs\yolov52\lib\site-packages\torch\serialization.py", line 773, in _legacy_load
result = unpickler.load()
ModuleNotFoundError: No module named 'mc.build.lib'

还是报错

[wuzhenxin1989]

@WongKinYiu 大佬看下

[WongKinYiu]

ModuleNotFoundError: No module named 'mc.build.lib'
這裡就是說您沒有mc/build/lib呀

[businiaoo]

when I run:
python setup.py build

Report an error：
ScaledYOLOv4-yolov4-large/mc/csrc/mish_kernel.cu:3:30: fatal error: CUDAApplyUtils.cuh: No such file or directory

Do you know how to solve it？

installation is updated.

Thank you. I made it

[WongKinYiu]

you could try https://github.com/JunnYu/mish-cuda

[BaofengZan]

@WongKinYiu 你好，导出onnx时出错：ONNX export failure: ONNX export failed: Couldn't export Python operator MishCudaFunction。这种还需要更改哪一块代码么？

[WongKinYiu]

這比較麻煩 得改用pytorch原生程式碼構建mish 但會比較慢也比較吃資源

class Mish(nn.Module):  # https://github.com/digantamisra98/Mish
    def forward(self, x):
        return x * F.softplus(x).tanh()