UtiliViz

System utilization visualizer.

A simple module to support statistic profiling of system utilization. Currently monitors memory and processor utilization.

Supports NVIDIA GPU via py3nvvml.

Dependencies

Required:

• Python 3+
• psutil

Optional:

• py3nvml is needed for profiling CUDA GPUs: pip install py3nvml

API

Use as contextmanager

In regular python, use record(list_of_monitors) as a context-manager to contain the code being profiled:

import utiliviz as uv

with uv.record([uv.CpuMon]) as rec:
    some_code()            # this will be profiled
raw_data = rec.get_data()  # get raw data
plots = rec.make_plot()    # get bokeh plots

Monitors

• The CPU monitor is utiliviz.CpuMon. It checks the system RAM and per-core utilization. The overall CPU utilization is also showed.

• The utiliviz.CpuOverallMon is a subset of utiliviz.CpuMon that only shows the overall CPU utilization.

• CUDA-GPU profiling is enabled by using the CudaGpuMon monitor class.

Use as magic

Provides IPython magic utiliviz for Jupyter notebook usage.

The magic must be registered with:

import utiliviz as uv
# setup_bokeh controls whether `bokeh.io.output_notebook()` is called.
uv.register_magic(setup_bokeh=True)

Magic cell command usage:

    %%utiliviz [--nocpu] [--cuda]