GeForce GTX 1080, CUDA 8.0, Ubuntu 16.04, Caffe
Pages 35
- Home
- AWS EC2 GPU enabled Caffe AMI
- Borrowing Weights from a Pretrained Network
- Caffe installing script for ubuntu 16.04 support Cuda 8
- Caffe on EC2 Ubuntu 14.04 Cuda 7
- Caffe Output: .caffemodel .solverstate
- Contributing
- Development
- Excluding Layers: Train and Test Phase
- Faster Caffe Training
- Fine Tuning or Training Certain Layers Exclusively
- GeForce GTX 1080, CUDA 8.0, Ubuntu 16.04, Caffe
- IDE Nvidia’s Eclipse Nsight
- Image Format: BGR not RGB
- Install Caffe on EC2 from scratch (Ubuntu, CUDA 7, cuDNN 3)
- Installation
- Installation (OSX)
- Making Prototxt Nets with Python
- Model Zo
- Model Zoo
- Models accuracy on ImageNet 2012 val
- OpenCV 3.2 Installation Guide on Ubuntu 16.04
- Python Layer Unit Tests
- Related Projects
- Reporting Bugs and Other Issues
- Simple Example: Sin Layer
- Solver Prototxt
- The Data Layer
- The Datum Object
- Training and Resuming
- Ubuntu 14.04 ec2 instance
- Ubuntu 14.04 VirtualBox VM
- Ubuntu 16.04 or 15.10 Installation Guide
- Using a Trained Network: Deploy
- Working with Blobs
- Show 20 more pages…
Clone this wiki locally
Install cuda 8.0.
As the default gcc in Ubuntu 16.04 is very new, if you have compiling error similar to #error -- unsupported GNU version! gcc versions later than 5.3 are not supported!
Try to comment the #error line in file /usr/local/cuda/include/host_config.h
#if __GNUC__ > 5 || (__GNUC__ == 5 && __GNUC_MINOR__ > 3)
//#error -- unsupported GNU version! gcc versions later than 5.3 are not supported!
#endif /* __GNUC__ > 5 || (__GNUC__ == 5 && __GNUC_MINOR__ > 1) */
Update nvidia driver
You may need to remove the old driver version installed with cuda-8.0 and install a newer version of driver, if you have error similar to modprobe: ERROR: could not insert 'nvidia_361_uvm': Invalid argument
Please refer here to remove old driver, then install a new one:
sudo apt-get purge nvidia-*
dkms status
do this based on your dkms status:
sudo dkms remove bbswitch/0.8 -k 4.4.0-31-generic
download the compatible driver (GTX 1080 in my case) from nvidia website and install it
sudo ./NVIDIA-Linux-x86_64-367.35.run
reboot if the driver was not updated
sudo reboot
compile the sample code and check if it works
./cuda8-0-samples/bin/x86_64/linux/release/deviceQuery
Install prerequisities of Caffe
Because the gcc version is Ubuntu 16.04 is very new, If any prerequisity installed from apt-get does not work, uninstall it, compile and install it by the default gcc (5.4) from the source code. The prerequisities that may have problems include: protobuf and opencv. E.g. if you have protobuf error similar to
.build_release/lib/libcaffe.so: undefined reference to 'google::protobuf::io::CodedOutputStream::WriteVarint64ToArray(unsigned long long, unsigned char*)'
Try to uninstall the protobuf installed by apt-get, the one installed by apt-get might have been compiled by a older gcc version, so its shared libraries may not be compatible with your default gcc:
sudo apt-get autoremove libprotobuf-dev protobuf-compiler
then compile the protobuf-2.5.0 from src and install it. Please config the default gcc (5.4 in my case) when you compile protobuf:
./configure --prefix=/your/path/ CC=/usr/bin/gcc
make
make check
make install