How can I get Caffe or NVCaffe installed on my DPX2

I’m trying to install nvcaffe with the below link, but errors occurred when I execute make all -j4

Here is the errors prompts:

CXX examples/siamese/convert_mnist_siamese_data.cpp
CXX .build_release/src/caffe/proto/
LD -o .build_release/lib/
AR -o .build_release/lib/libcaffe-nv.a
/usr/bin/ld: /usr/lib/gcc/aarch64-linux-gnu/5/../../../aarch64-linux-gnu/libturbojpeg.a(libturbojpeg_la-turbojpeg.o): relocation R_AARCH64_ADR_PREL_PG_HI21 against external symbol `__stack_chk_guard@@GLIBC_2.17' can not be used when making a shared object; recompile with -fPIC
/usr/bin/ld: /usr/lib/gcc/aarch64-linux-gnu/5/../../../aarch64-linux-gnu/libturbojpeg.a(libturbojpeg_la-turbojpeg.o)(.text+0x64): unresolvable R_AARCH64_ADR_PREL_PG_HI21 relocation against symbol `__stack_chk_guard@@GLIBC_2.17'
/usr/bin/ld: final link failed: Bad value
collect2: error: ld returned 1 exit status
Makefile:600: recipe for target '.build_release/lib/' failed
make: *** [.build_release/lib/] Error 1

Is there a guideline to install caffe or NVcaffe on px2? thank you.

Dear zew_wang,

Can I know why you want to install Caffe or NVcaffe on DPX2?
Because DPX2 is a platform suitable for deployment rather than training. Therefore, it is a good system to use the trained model to refer to using TensorRT. Thanks.

Dear Steve

I want to train the model with caffe, since training on the DPX2 will take fewer time than on my pc.

Does PX2 support any deep learning framework that I can training on? Thanks.

Dear zew_wang,

The link below may help. Thanks.

Hi Steve, I have got caffe installed on dpx2 and I’m training a model base on VOC dataset.

The training process is quite slow, even slower than my pc(GTX 1070)

When I set the batchsize to 8, it will out of memory.

I’m curious about that, Why?

Dear zew_wang,

The training process is quite slow, even slower than my pc(GTX 1070)
I think this is expected behavior. Please check GTX1070 spec and dGPU spec on DPX2.
GTX 1070 :
dGPU in DPX2 : ./deviceQuery in /NVIDIA_CUDA-9.0_Samples/1_Utilities/deviceQuery$ on DPX2. Thanks.