Skip to content

UnaryLab/chopper

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

27 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Chopper - GPU Characterization Tool

A GPU characterization tool that takes chrome traces and optionally hardware counters for multiple GPUs as input, and generates compute, memory, and architecture visualizations.

Setup

To install chopper run:

pip install .

Profiling

The chopper profiler is invoked using its python module. In this example, GPU and CPU telemetry are enabled.

python -m chopper.profile.collect --gpu-telemetry --cpu-telemetry -- <workload args>

It is up to the workload to enable the PyTorch profiler.

To only execute the workload args and/or hardware profiler on a container while collecting telemetry outside of it, pass --container with the apptainer .sif file path. Otherwise, chopper can be installed and ran inside a container.

Visualization

To select and view plots run:

python -m chopper.window

Tests

To type check with mypy run:

python -m tests

Related projects

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages