Kevin Tan

使用 VS 2022 打开 CMake 生成的 MNN.sln，选择“生成解决方案”，然后出现相同问题： ``` 12>MNN_CUDA.lib(MNN_CUDA_generated_Int8ToFloatExecution.cu.obj) : error LNK2038: 检测到“RuntimeLibrary”的不匹配项: 值“MT_StaticRelease”不匹配值“MD_DynamicRelease”(Backend.obj 中) 12>MNN_CUDA.lib(MNN_CUDA_generated_FloatToInt8Execution.cu.obj) : error LNK2038: 检测到“RuntimeLibrary”的不匹配项: 值“MT_StaticRelease”不匹配值“MD_DynamicRelease”(Backend.obj 中) 12>MNN_CUDA.lib(MNN_CUDA_generated_DepthwiseConvInt8Execution.cu.obj) : error LNK2038: 检测到“RuntimeLibrary”的不匹配项: 值“MT_StaticRelease”不匹配值“MD_DynamicRelease”(Backend.obj 中) 12>MNN_CUDA.lib(MNN_CUDA_generated_CutlassGemmInt8TensorCore16832.cu.obj)...

使代码更简洁，同时避免因使用-O类似选项启动Python解释器时不编译assert语句而产生的问题。在 2022年12月22日，20:06，Bear Huchao ***@***.***> 写道： @bearbattle commented on this pull request. 这样做有什么好处吗？ —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the...

It is because of `#define int long long`. Change `int main(int argc, char **argv)` to `signed main(signed argc, char **argv)` works.

Thanks a lot! I realized that when I inspected the type enum. But the MinGW 8.1.0 also cannot compile it if `argc` has the type `long`. Maybe I can modify...

你应该能看到sciter

I think it's slow partly because some tensor operations are done on the CPU rather than the GPU, so all physical CPU cores are consumed throughout the entire training process....

Will you be releasing a version of this bug fix in the near future, as it is indeed a serious bug?

Thank you, then I’ll wait for your new build instead of building my own version or reverting to the old one:)