HDF5DataLayer with GPU

[mjj...@gmail.com]

System configuration

• Operating system: ubuntu14.04
• Compiler: gcc4.8
• CUDA version (if applicable): 8.0
• CUDNN version (if applicable): 6.0
• Python version (if using pycaffe):2.7

Hi,all. When I try to use HDF5DataLayer,I found it works right in cpu mode, but works wrong in gpu mode.
my net like bellow:

layer { name: "data" type: "HDF5Data" top: "data" top: "label" hdf5_data_param { source: "train_h5_list.txt" batch_size: 1 } } layer { name: "conv1_1" type: "Convolution" bottom: "data" top: "conv1_1" param { lr_mult: 1.0 decay_mult: 0.0 } param { lr_mult: 1.0 decay_mult: 0.0 } convolution_param { num_output: 48 pad: 1 kernel_size: 3 stride: 2 weight_filler { type: "xavier" } bias_filler { type: "constant" value: 0.0 } } }

the strange is, there is no error just breakdown, like bellow:

I0509 15:34:06.122910 15842 net.cpp:255] Network initialization done. I0509 15:34:06.122928 15842 solver.cpp:57] Solver scaffolding done.




are there anyone met the same problem like me?thx very much!

[mjj...@gmail.com]

OK~ the problem solved by pulling the newest caffe version

在 2018年5月9日星期三 UTC+8下午3:35:52，mjj...@gmail.com写道：