-
-
Save jetsonhacks/acf63b993b44e1fb9528 to your computer and use it in GitHub Desktop.
#!/bin/sh | |
# Install and compile Caffe on NVIDIA Jetson TK1 Development Kit | |
sudo add-apt-repository universe | |
sudo apt-get update | |
sudo apt-get install libprotobuf-dev protobuf-compiler gfortran \ | |
libboost-dev cmake libleveldb-dev libsnappy-dev \ | |
libboost-thread-dev libboost-system-dev \ | |
libatlas-base-dev libhdf5-serial-dev libgflags-dev \ | |
libgoogle-glog-dev liblmdb-dev -y | |
sudo usermod -a -G video $USER | |
# Git clone Caffe | |
sudo apt-get install -y git | |
git clone https://github.com/BVLC/caffe.git | |
cd caffe && git checkout dev | |
cp Makefile.config.example Makefile.config | |
make -j 4 all | |
make -j 4 runtest | |
build/tools/caffe time --model=models/bvlc_alexnet/deploy.prototxt --gpu=0 |
Jetson TK1 can use only Cuda 6.5 so it means you can use only cuDNN v2.
There is no branch or tag named dev, since getting following error
error: pathspec 'dev' did not match any file(s) known to git.
Ok finally was able to build the caffe , but not the dev branch, even dev branch does not exist now, They have merged it to master , and master can be only compiled using cuDNN 5+ which require CUDA 7, but currently for Jetson TK1 latest CUDA version is 6.5 and cuDNN last supported version for that is cudnn-6.5-linux-ARMv7-v2.So can't build the master branch in TK1.
Instead, I was able to build this fork which is for cudnn 2.x
I had same error: pathspec 'dev' ... Currently the script is in a random test phase and seems to be passing the tests. What problem am I going to have?
able to build using master
branch
can you help me to install caffe on ubunto 16.04 please :)
I had flash tk1 by JetPack first,Then,I run this script, I got some error, the first one is "there is no 'dev' branch " same with tmkasun.
I maybe get the key. To make the master of caffe for tk1, we should comment compute_60 in makefile.config .just easy
This one works for CUDNN v2 on Jetson TK1. https://github.com/platotek/caffetk1. Also make sure to install g++4.8 before you run the make commands.
I just wrote a GIST on this issue.
Jetson TK1 - caffe compilation issue: nvcc fatal : Unsupported gpu architecture 'compute_60'
In short, download caffe rc2 from releases and compile using this build. You need CUDA 9 and yes it sucks that Kepler GPUs are supported just not this one ( Kepler "GK20a" GPU ) and the last Jetpack release had no updates for TK1.
mada some mistakes, because of the version of caffe and cuDNN