torchtel 0.0.2

OpenTelemetry auto instrumentation for pytorch

torchtel

torchtel provides OpenTelemetry auto instrumentation for pytorch

Requirements

• Python 3.10 or higher
• PyTorch
• OpenTelemetry API
• OpenTelemetry Instrumentation ≥0.56b0

Install from pypi

$ pip install torchtel

Usage

Instrument PyTorch

To instrument PyTorch, you need to import the PyTorchInstrumentor class, call the instrument() method, and register your model, see below:

from torchtel import PyTorchInstrumentor

# create your PyTorch model
model = ...

# Registed the model for PyTorch Instrumentation
instrumentor = PyTorchInstrumentor().instrument(model=model)

# Run the model
...

# Uninstrument PyTorch
instrumentor.uninstrument()

Check out the model example to see OpenTelemetry autoinstrumentation in action.

Observability

To visualize traces and metrics collected by TorchTel, see our complete observability setup guide with Docker Compose configurations for Jaeger, Prometheus, and Grafana.

Contributing

Read our contributing guide to learn about our development process, how to propose bugfixes and feature requests, and how to build your changes.

Code of Conduct

Facebook has adopted a Code of Conduct that we expect project participants to adhere to. Please read the full text so that you can understand what actions will and will not be tolerated.

Mantainers

clusterscope is actively maintained by Lucca Bertoncini, and Billy Campoli.

License

clusterscope is licensed under the MIT license.