openeuler/caffe repository overview

⁠Quick reference

• The official Caffe docker image.

• Maintained by: openEuler CloudNative SIG⁠.

• Where to get help: openEuler CloudNative SIG⁠, openEuler⁠.

⁠Caffe | openEuler

Current Caffe docker images are built on the openEuler⁠. This repository is free to use and exempted from per-user rate limits.

Caffe is a deep learning framework made with expression, speed, and modularity in mind.

Learn more about on Caffe Website⁠.

⁠Supported tags and respective Dockerfile links

The tag of each caffe docker image is consist of the complete software stack version. The details are as follows

⁠Usage

In this usage, users can select the corresponding {Tag} and container startup options based on their requirements.

• Pull the openeuler/caffe image from docker

  docker pull openeuler/caffe:{Tag}
  

  Copy

• Run with an interactive shell

  You can also start the container with an interactive shell to use caffe.

  docker run -it --rm openeuler/caffe:{Tag} bash
  

  Copy

• How to train LeNet on MNIST with BVLC Caffe (CPU/GPU)

  • Download the MNIST dataset manually(mirror)

    In your container, go to mnist directory:

    cd /opt/caffe/data/mnist
    

    Copy

    Download the files from a reliable public mirror:

    wget https://ossci-datasets.s3.amazonaws.com/mnist/train-images-idx3-ubyte.gz
    wget https://ossci-datasets.s3.amazonaws.com/mnist/train-labels-idx1-ubyte.gz
    wget https://ossci-datasets.s3.amazonaws.com/mnist/t10k-images-idx3-ubyte.gz
    wget https://ossci-datasets.s3.amazonaws.com/mnist/t10k-labels-idx1-ubyte.gz
    

    Copy

    Uncompress the files:

    gzip -d *.gz
    

    Copy

  • Convert the dataset to LMDB

    Switch to the Caffe root directory and run:

    cd /opt/caffe
    ./examples/mnist/create_mnist.sh
    

    Copy

    This script converts the raw MNIST files to LMDB format which Caffe uses for training.

  • Check the solver configuration

    Open the solver prototxt file:

    vi ./examples/mnist/lenet_solver.prototxt
    

    Copy

    Make sure the last line matches your build:

    solver_mode: CPU   # For CPU-only Caffe builds
    

    Copy

    or

    solver_mode: GPU   # For Caffe built with CUDA/cuDNN
    

    Copy

  • Train the model

    Finally, run the training command:

    ./build/tools/caffe train --solver=examples/mnist/lenet_solver.prototxt
    

    Copy

    Caffe will start training the LeNet model on the MNIST dataset.

⁠Question and answering

If you have any questions or want to use some special features, please submit an issue or a pull request on openeuler-docker-images⁠.