mmchen88

Results 10 comments of mmchen88

same problem

ONNX export failure: Tensors must have same number of dimensions: got 1 and 2

> You can replace mmskeleton/models/backbones/st_gcn_aaai18.py line104 with: x = F.avg_pool2d(x, [75,18]) ONNX export failure: Tensors must have same number of dimensions: got 1 and 2

when opt.hybridize=True x in the hybrid_forward(self, F, x) is a Symbol data, while when opt.hybridize=False x in the hybrid_forward(self, F, x) is a NDArray

BN_inf_Fwd函数里应该修改这一句: Dtype scale_factor = factor_b->cpu_data()[0] == 0 ? 0 : (1 / factor_b->cpu_data()[0]); 如下: Dtype scale_factor = factor_b->cpu_data()[0] == 0 ? 1 : (1 / factor_b->cpu_data()[0]);

> > 请问CPU_ONLY怎么开启啊,我用的CPU模式控制台一闪而过就没了 > > 您好,请问您解决了吗?我想只用CPU编译caffe,但博主好像没有提供CommonSetting文件,不知道在哪里配置不用GPU的选项。 工程属性里添加CPU_ONLY的宏定义就行了吧,在C/C++->预处理器定义里

> > > > 请问CPU_ONLY怎么开启啊,我用的CPU模式控制台一闪而过就没了 > > > > > > > > > 您好,请问您解决了吗?我想只用CPU编译caffe,但博主好像没有提供CommonSetting文件,不知道在哪里配置不用GPU的选项。 > > > > > > 工程属性里添加CPU_ONLY的宏定义就行了吧,在C/C++->预处理器定义里 > > 我添加了宏定义,但编译的时候还是会去编译.cu,然后就报错 那就将cu文件移除工程,把对应的cpp添加进工程,再编译

Same question. After training the ucf101 model, I use the method "CopyTrainedLayersFrom" to parse the caffemodel file, but it failed. There are only 2 layers saved in the caffemodel file:...

install RapidJSON https://blog.csdn.net/lj19940110/article/details/101012837