Description
Agent Platform tuning is an agent skill for fine-tuning large language models on Google Cloud Agent Platform, for both open models and Gemini models. It describes the whole lifecycle of a tuning project, from setting up the environment and preparing data, through job configuration and monitoring, to deploying the tuned model. A workflow decision tree guides the agent through the choices. The first check is whether you have provided the Google Cloud project and region, since those are needed before anything runs. From there the skill branches according to the model type and the stage you are at. The skill is intended for fine-tuning with the Agent Platform tuning service, and it says not to use it for training models from scratch, where another route applies. It also points to a separate skill for managing tuning jobs once they are running. It is useful for teams that have a dataset of examples and want to adapt a model to their domain on managed infrastructure. It is published by Google on skills.sh under the google skills repository, with about 21.3K stars and about 8.2K installs. The skills.sh listing shows Gen Agent Trust Hub and Socket passing, and Snyk with a Warn result.…
Highlights
- Tunes open and Gemini models
- Covers data preparation to deployment
- Decision tree for each step
- Checks project and region first
- Points to a separate job-management skill
Ratings & Reviews
No ratings yet
Be the first to rate agent-platform-tuning.
People also visit
agent-platform-eval-flywheel
Measures and improves the quality of AI models and agents on Google Cloud
agent-platform-rag-engine-management
Manage and query Agent Platform RAG Engine Corpora
agent-platform-troubleshooting
Troubleshoots Google Cloud Gemini Enterprise Agent Platform issues
agent-platform-tuning FAQ
What does the skill cover?
Fine-tuning open models and Gemini models with the Agent Platform tuning service, from environment setup to deployment.
What does it need first?
The Google Cloud project and region. The skill checks for these before starting.
Is it for training from scratch?
No. The skill says not to use it for model training, which needs a different route.
How do I manage a running job?
The separate tuning management skill lists, gets and cancels tuning jobs.
How do I install it?
Run npx skills add https://github.com/google/skills --skill agent-platform-tuning.
What are the best agent-platform-tuning alternatives?
Popular alternatives to agent-platform-tuning include agent-platform-inference, agent-platform-deploy and agent-platform-eval-flywheel. See all agent-platform-tuning alternatives