gemma_domain_trainer_prototype - Fine-tune Gemma with domain-specific LoRA data
Fine-tunes Gemma 270M with domain-specific instruction data using LoRA, then validates the model and generates deployment configuration.
Tags
Updated: 2026-10-01Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Load and validate datasets
- Format Gemma training examples
- Configure LoRA hyperparameters
- Train LoRA adapters
- Validate model outputs
- Generate deployment configuration
- Log execution patterns
Inputs
- Domain training dataset path
- Knowledge domain
- Training parameters
- Base Gemma model path
Outputs
- Fine-tuned LoRA adapter files
- Training metrics JSON file
- Execution identifier
- Breadcrumb execution logs
- Deployment configuration
Requirements
- Gemma 270M base model
- LoRA training support
- Transformers and PEFT libraries
- Read access to training data
- Write access to model and metrics paths