unsloth-inference - Optimized inference deployment for fine-tuned models
Deploy fine-tuned models for production inference using kernel optimization, vLLM, or SGLang
Tags
Updated: 2026-02-15Capabilities
Typical Inputs
Typical Outputs
What this skill does
- enable optimized kernels
- merge LoRA weights
- export for production serving
- provide OpenAI-compatible endpoint
Inputs
- fine-tuned model
- tokenizer
- LoRA weights
Outputs
- optimized inference model
- merged model files
- API endpoint
Requirements
- unsloth library
- PyTorch
- vLLM (optional)
- SGLang (optional)