Skip to main content

​
Overview

This guide shows you how to integrate Latitude Telemetry into an application that runs models locally with Hugging Face Transformers.
You’ll keep calling Transformers exactly as you do today. Telemetry simply observes and enriches those calls.
Auto-instrumentation for Transformers is available in the Python SDK. For other languages, send traces via the OpenTelemetry exporter.

​
Requirements

  • A Latitude account and API key
  • A Latitude project slug
  • A project that uses the Transformers library (transformers)
  • A Hugging Face token for gated models (set as HF_TOKEN)

​
Steps

1

Install

pip install latitude-telemetry transformers torch
2

Initialize and use

import transformers
from transformers import pipeline

from latitude_telemetry import Latitude, capture

latitude = Latitude(
    api_key="your-api-key",
    project="your-project-slug",
    instrumentations={"transformers": transformers},
)

def generate_reply():
    generator = pipeline("text-generation", model="gpt2", max_new_tokens=50)
    result = generator("Hello:")
    return result[0]["generated_text"]

capture("generate-reply", generate_reply)

latitude.shutdown()

​
Seeing Your Traces

Once connected, traces appear automatically in Latitude:
  1. Open your project in the Latitude dashboard
  2. Each execution shows input/output messages, model, token usage, latency, and errors