.\"Text automatically generated by txt2man .TH CAFFE 1 "10 August 2016" "" "" .SH NAME \fBcaffe \fP- command line brew for Caffe .SH SYNOPSIS .nf .fam C \fBcaffe\fP .fam T .fi .fam T .fi .SH 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. .SH COMMANDS .TP .B train train or finetune a model .TP .B test score a model .TP .B device_query show GPU diagnostic information .TP .B time benchmark model execution time .SH FREQUENTLY USED FLAGS .TP .B \fB-gpu\fP (Optional; run in GPU mode on given device IDs separated by ','. Use '\fB-gpu\fP all' to run on all available GPUs. The effective training batch size is multiplied by the number of devices.) type: string default: "" .TP .B \fB-iterations\fP (The number of iterations to run.) type: int32 default: 50 .TP .B \fB-level\fP (Optional; network level.) type: int32 default: 0 .TP .B \fB-model\fP (The model definition protocol buffer text file..) type: string default: "" .TP .B \fB-phase\fP (Optional; network phase (TRAIN or TEST). Only used for 'time'.) type: string default: "" .TP .B \fB-sighup_effect\fP (Optional; action to take when a SIGHUP signal is received: snapshot, stop or none.) type: string default: "snapshot" .TP .B \fB-sigint_effect\fP (Optional; action to take when a SIGINT signal is received: snapshot, stop or none.) type: string default: "stop" .TP .B \fB-snapshot\fP (Optional; the snapshot solver state to resume training.) type: string default: "" .TP .B \fB-solver\fP (The solver definition protocol buffer text file.) type: string default: "" .TP .B \fB-stage\fP (Optional; network stages (not to be confused with phase), separated by ','.) type: string default: "" .TP .B \fB-weights\fP (Optional; the pretrained weights to initialize finetuning, separated by ','. Cannot be set simultaneously with snapshot.) type: string default: "" .TP .B \fB-help\fP Show complete help messages. .SH OTHER CAFFE UTILITIES Apart from the "\fBcaffe\fP" command line utility, there are also some utilities available, run them with "\fB-h\fP" or "\fB--help\fP" argument to see corresponding help. .IP \(bu 3 convert_imageset .IP \(bu 3 convert_cifar_data .IP \(bu 3 compute_image_mean .IP \(bu 3 convert_mnist_siamese_data .IP \(bu 3 upgrade_net_proto_binary .IP \(bu 3 extract_features .IP \(bu 3 upgrade_solver_proto_text .IP \(bu 3 classification .IP \(bu 3 upgrade_net_proto_text .IP \(bu 3 convert_mnist_data .SH EXAMPLES Train a new Network .PP .nf .fam C $ caffe train -solver solver.prototxt .fam T .fi Resume training a network from a snapshot .PP .nf .fam C $ caffe train -solver solver.prototxt -snapshot bvlc_alexnet.solverstate .fam T .fi Fine-tune a network .PP .nf .fam C $ caffe train -solver solver.prototxt -weights pre_trained.caffemodel .fam T .fi Test (evaluate) a trained model for 100 iterations, on GPU 0 .PP .nf .fam C $ caffe test -model train_val.prototxt -weights bvlc_alexnet.caffemodel -gpu 0 -iterations 100 .fam T .fi Run a benchmark against AlexNet on GPU 0 .PP .nf .fam C $ caffe time -model deploy.prototxt -gpu 0 .fam T .fi Check CUDA device availability of GPU 0 .PP .nf .fam C $ caffe device_query -gpu 0 .fam T .fi .SH HOMEPAGE http://caffe.berkeleyvision.org .SH BUGS https://github.com/BVLC/\fBcaffe\fP/issues .SH AUTHOR This manpage is written by Zhou Mo with the help of txt2man for Debian according to program's help message.