Zer0Fit Serves Google's TabFM/TimesFM Zero-Shot

AI-Agents OpenSource

TL;DR: A grad student developed Zer0Fit, an MCP server that makes Google's new TabFM and TimesFM foundation models available for zero-shot ML tasks via a single Docker container.

Summary: Zer0Fit wraps Google's recently released TabFM and TimesFM transformer models, enabling zero-shot machine learning tasks. This PyTorch-based server, packaged in a Docker container, allows connection to local LLMs and achieves solid accuracy on classic ML datasets without traditional model training. It requires 16GB+ VRAM and supports CSV input, with more formats planned.

Why it matters: This tool simplifies access to powerful tabular and time-series foundation models, allowing indie developers to integrate advanced ML capabilities into their GenAI applications with minimal effort. Experiment with Zer0Fit to explore zero-shot ML for data analysis and prediction tasks, potentially accelerating feature development.

Source: reddit