Scroll to navigation

CAFFE(1) CAFFE(1)

NAME

caffe - command line brew for Caffe

SYNOPSIS


caffe <COMMAND> <FLAGS>

DESCRIPTION

Caffe is a deep learning framework made with expression, speed, and modularity in mind. It is developed by the Berkeley Vision and Learning Center (BVLC) and community contributors.

COMMANDS

train or finetune a model
score a model
show GPU diagnostic information
benchmark model execution time

FREQUENTLY USED FLAGS

(Optional; run in GPU mode on given device IDs separated by ','. Use '-gpu all' to run on all available GPUs. The effective training batch size is multiplied by the number of devices.) type: string default: ""
(The number of iterations to run.) type: int32 default: 50
(Optional; network level.) type: int32 default: 0
(The model definition protocol buffer text file..) type: string default: ""
(Optional; network phase (TRAIN or TEST). Only used for 'time'.) type: string default: ""
(Optional; action to take when a SIGHUP signal is received: snapshot, stop or none.) type: string default: "snapshot"
(Optional; action to take when a SIGINT signal is received: snapshot, stop or none.) type: string default: "stop"
(Optional; the snapshot solver state to resume training.) type: string default: ""
(The solver definition protocol buffer text file.) type: string default: ""
(Optional; network stages (not to be confused with phase), separated by ','.) type: string default: ""
(Optional; the pretrained weights to initialize finetuning, separated by ','. Cannot be set simultaneously with snapshot.) type: string default: ""
Show complete help messages.

OTHER CAFFE UTILITIES

Apart from the "caffe" command line utility, there are also some utilities available, run them with "-h" or "--help" argument to see corresponding help.

  • convert_imageset
  • convert_cifar_data
  • compute_image_mean
  • convert_mnist_siamese_data
  • upgrade_net_proto_binary
  • extract_features
  • upgrade_solver_proto_text
  • classification
  • upgrade_net_proto_text
  • convert_mnist_data

EXAMPLES

Train a new Network


$ caffe train -solver solver.prototxt
Resume training a network from a snapshot


$ caffe train -solver solver.prototxt -snapshot bvlc_alexnet.solverstate
Fine-tune a network


$ caffe train -solver solver.prototxt -weights pre_trained.caffemodel
Test (evaluate) a trained model for 100 iterations, on GPU 0


$ caffe test -model train_val.prototxt -weights bvlc_alexnet.caffemodel -gpu 0 -iterations 100
Run a benchmark against AlexNet on GPU 0


$ caffe time -model deploy.prototxt -gpu 0
Check CUDA device availability of GPU 0


$ caffe device_query -gpu 0

HOMEPAGE

http://caffe.berkeleyvision.org

BUGS

https://github.com/BVLC/caffe/issues

AUTHOR

This manpage is written by Zhou Mo <cdluminate@gmail.com> with the help of txt2man for Debian according to program's help message.

10 August 2016