OpenHands and Snowflake launch Agent harness and training workflow
The story so far
OpenHands has partnered with Snowflake AI Research to introduce a workflow that allows developers to directly use the existing OpenHands Agent harness for reinforcement learning post-training, without rebuilding the harness in the training framework. According to OpenHands, the local machine only needs to run agent rollout on the CPU, with model training and sampling completed on Snowflake Cortex Training GPUs, each defaulting to four H200s. One run requires only one Snowflake account with Cortex Training enabled and a set of Harbor tasks. There is currently one limitation to this workflow: the OpenHands file_editor tool is unavailable during training, and agents must edit files via the terminal.
AI-written from the reports · Updated 14 days ago
Coverage timeline
Follow the coverage to see every side of the story.
- OpenHands BlogFeaturedPost-Train a Model for Your Own Agent Harness with OpenHands and Snowflake
OpenHands partnered with Snowflake AI Research, allowing developers to directly use the existing OpenHands Agent harness for reinforcement learning and training without rebuilding the harness in the training framework.