The last essay talked about the process of installing caffe on windows and running the mnist routine. Seriously, just like a nurse, every step has to give milk to work. Recently, a desktop computer has been configured, which can be used as a reference for configuring the learning machine in the future.
The configuration is as follows: Supplementary image.
The installation process is as follows:
1. nvidia driver
2. cuda, download cuda 8.0 on the official website, I downloaded the .deb file, install it according to the instructions on the official website, and finally need to add some content to the .bashrc file, the content I add here can go to the install guide Found, can also be found and added through other people's blogs.
3. For the installation of opencv 3.3, you can refer to my previous blog. I originally installed the opencv 3.1 version in advance this time. Later, I could not install it successfully after compiling many times, so I decisively changed the 3.3 version and compiled it once. , installation, it seems that sometimes installing software needs to consider the version, not the older the more stable the version, the easier it is to install.
4. The installation of caffe is still carried out according to the instructions on the official website + my previous blog. During the installation process, the mnist training routine was run, and it was found that it was completed in seconds. When executing ./train_lenet.sh, I found that there are some problems in execution. The sudo ./train_lenet.sh command suggested on the blog can solve this problem, and the suggestion on the blog to reinstall the 1080 ti driver is quite tricky. It took some time to install this driver, and reinstalling his is impossible.
5. After tossing around for a long time, I used bazel to compile TensorFlow. Unfortunately, I always made mistakes. I found that it was impossible to install it with the help of pip and pip3 commands. The brothers and sisters easily completed the installation with Anaconda on this computer. , then use the brothers and sisters to install it.
6. In a few days, I need to install mxnet on ubuntu. After all, Mr. Li Mu provided a set of very powerful deep learning tutorials, emphasizing practice.