Training Dashboards with Trackio + Hugging Face

2026-01-22 · Source: HuggingFace · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Intermediate, quick

Summary

The content describes how Trackio integrates with the Hugging Face Hub to visualize training metrics directly on a model's page. This integration allows users to fine-tune models using Hugging Face Transformers and Trainers, with a specific `report to tracheo` line automatically logging metrics to a local Trackio project. When the model is pushed to the Hugging Face Hub, the associated Trackio data, including logs and datasets, is simultaneously pushed to a dedicated Hugging Face Space under the user's namespace. This process requires Transformers version 5.0 or higher. Once pushed, a "Visualize and Trackio" badge appears on the model page, and a new "Training Metrics" tab displays the Trackio dashboard, showing all logged metrics, including system metrics and media tables, directly within the Hugging Face interface.

Key takeaway

For MLOps Engineers managing model deployments, integrating Trackio with Hugging Face Hub streamlines metric visualization. This setup allows you to centralize training performance data directly on your model cards, simplifying monitoring and collaboration without needing separate dashboards. You should update your Transformers library to version 5.0 or higher to enable this direct integration.

Key insights

Trackio integrates with Hugging Face Hub to display training metrics directly on model pages.

Principles

• Automate metric logging during model training.
• Centralize model and training data on the Hub.

Method

Fine-tune a model with Hugging Face Transformers, add `report to tracheo` for metric logging, then push the model to the Hugging Face Hub, which automatically pushes Trackio data to a Space.

In practice

• Use `report to tracheo` in training scripts.
• Ensure Transformers 5.0+ is installed.
• Check the "Training Metrics" tab on Hugging Face.

Topics

• Hugging Face Hub
• Model Training
• Training Metrics
• Transformers Library
• Sentiment Classification

Best for: Machine Learning Engineer, MLOps Engineer, AI Engineer

Related on AIssential

• How to Actually Get Started with HuggingFace 🤗 — Hugging Face provides an open-source ecosystem for ML/NLP, simplifying model access, development, and deployment.
• Trackio Tutorial: Hugging Face's new, FREE experiment tracking library — Track You offers local-first, open-source ML experiment tracking with W&B-compatible API and AI agent integration.
• 📊 ArtifactLinker: a GNN ranks which HuggingFace models will hit SOTA on which benchmarks; — ArtifactLinker uses a GNN to predict SOTA HuggingFace model performance on benchmarks.
• DeepInfra on Hugging Face Inference Providers 🔥 — DeepInfra's integration with Hugging Face Hub streamlines serverless AI inference for diverse models.
• What is Hugging Face? - Models, Datasets & Spaces — Hugging Face provides a unified platform for ML models, datasets, and application deployment.
• H2O MLOps Enterprise Model Registry & Hugging Face | Part 8 Integration | Part 8 — A robust model repository is essential for managing and governing a large, diverse portfolio of AI models in production.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by HuggingFace.