12. Command line examples

View this HTML help:

python3 -m veles --html-help

View help in terminal:

python3 -m veles --help

Interactively compose the command line in web browser and run it:

python3 -m veles --frontend

To run the Model, set path to the workflow and path to the configuration file (replace <workflow> and <config> with actual workflow and configuration paths):

python3 -m veles <workflow> <config>

Run “Mnist” sample:

python3 -m veles -s veles/znicz/samples/MNIST/mnist.py -


“-” is a shorthand for veles/znicz/samples/MNIST/mnist_config.py.


If you see warnings “Launcher:Failed to upload the status”, use “-s” option to disable reporting status to the Web Status Server.

Specify OpenCL device at once:

python3 -m veles -d 0:0 <workflow> <config>


The first number is the platform number and the second is the device number in that platform, for example “NVIDIA” platform and first (probably, only) device.

Specify CUDA device at once:

python3 -m veles -d 0 <workflow> <config>

Change backend (“auto” for AutoDevice, “cuda” for CUDADevice, “numpy” for NumpyDevice, “ocl” for OpenCLDevice):

python3 -m veles --backend "cuda" <workflow> <config>

Training from snapshotted state of the model (replace <snapshot> with actual path to the snapshot, could be amazon link):

python3 -m veles --snapshot <snapshot> <workflow> <config>

Run in testing mode (use trained model):

python3 -m veles --test --snapshot <snapshot> <workflow> <config>

Run workflow without any OpenCL/CUDA/any accelerations usage (slooow!):

python3 -m veles --force-numpy <workflow> <config>

Disable plotters during workflow run:

python3 -m veles -p '' <workflow> <config>

Do not send reports to web status server (for example, because it is not running):

python3 -m veles -s <workflow> <config>

Write only warnings and errors:

python3 -m veles -v warning <workflow> <config>

Write extended information from unit of class “Class”:

python3 -m veles --debug Class <workflow> <config>

Draw specific workflow’s scheme:

python3 -m veles --workflow-graph scheme.png <workflow> <config>
xdg-open scheme.png

12.1. Distributed training. Command line examples

Run workflow in distributed environment:

# on master node
python3 -m veles -l <workflow> <config>
# on slave node
python3 -m veles -m <master host name or IP>:5000 <workflow> <config>


5000 is the port number - use any you like!

Run workflow in distributed environment (known nodes):

# on master node
python3 -m veles -l -n <slave 1>/cuda:0,<slave 2>/ocl:0:1 <workflow> <config>


It’s ok to use different backends to the each slave node. “ocl:0:1” sets the OpenCL device to use. “cuda:0” sets the CUDA device to use. Syntax can be much more complicated, for example, cuda:0-3x2 launches 8 instances overall: two instances on each device from 0, 1, 2 and 3.

Table Of Contents

Previous topic

11. Installing PyPy (Ubuntu)

Next topic

13. How to use Veles. MNIST example. Simple (user entry/medium level)

This Page