Weights & Biases Unveils New Suite of Tools to Support LLMOps
SAN FRANCISCO, April 20, 2023 — Weights & Biases today announced a new suite of LLMOps tools to support Prompt Engineers, unlocking the true potential of large language models (LLMs) for massive business impact. The new suite of tools, W&B Prompts, will allow machine learning practitioners to build state-of-the-art LLM-based applications with speed and intelligently iterate on chains of prompts in real-time to find the inputs that lead to key insights.
The release of GPT-4 and ChatGPT has sparked an avalanche of interest and investment in LLMs and their potential, and the applications in areas like search engines, NLP, healthcare, robotics, and code generation are limitless. Despite this, few businesses are taking advantage of LLMs in a cohesive manner to solve real, tangible problems. LLM use cases are complex and require domain expertise, rapid, continuous iteration, and significant compute resources.
The companies developing the world’s most sophisticated LLMs – including OpenAI, Stability AI and Cohere – rely on Weights & Biases as a core part of their machine learning workflow. Building on the existing product that supports creating LLMs from scratch and fine-tuning them, the latest LLM-focused enhancements will increase productivity and collaboration for ML practitioners and Prompt Engineers.
“Weights & Biases is known for helping the world’s most sophisticated organizations build large language models, and with the introduction of W&B Prompts, we’re opening up a whole new way for businesses to harness the potential of LLMs,” said Phil Gurbacki, VP Product at Weights & Biases. “Regardless of how you’re working with LLMs, the Weights & Biases’ platform tracks activity to provide users with the interactive analysis required for iterative development of LLMs and LLM powered applications.”
The W&B Prompts suite of tools expands Weights & Biases’ best-in-class scalable experiment tracking, and are specifically designed with the needs of Prompt Engineers in mind. The new features and enhancements include:
- Tools for LLM debugging – Trace Timeline, Trace Table and Model Architecture expand experiment tracking from prompt engineers to easily review past results, identify and debug errors, glean results about model behavior, and share learnings with colleagues.
- W&B Launch to OpenAI Evals – new integrated framework for W&B Launch evaluation jobs and an easy, one-click way to evaluate and score prompts through the OpenAI Evals framework directly from the W&B UI.
- W&B Tables Enhancements – UX improvements within W&B Tables designed to provide better support for analysis and visualizations for Prompt Engineers working on LLMs.
W&B Prompts is now available to all Weights & Biases users. Visit the website to learn more.
About Weights & Biases
Weights & Biases is the leading developer-first MLOps platform that provides enterprise-grade, end-to-end MLOps workflow to accelerate ML activities. Used by over 500k ML practitioners including teams at OpenAI, Lyft, Blue River Technology, and Toyota, Weights & Biases is part of the new standard of best practices for machine learning.
Source: Weights & Biases