last updated August 1st 2024

About

Trelis Research provides:

• Advanced LLM fine-tuning scripts.
• Advanced LLM inference guides.
• Advanced vision fine-tuning scripts.
• Advanced voice/transcription fine-tuning scripts.
• One click LLM API templates
• Function-calling models.

ADVANCED Fine-tuning Scripts

Get access here

• retrieval (bm25, similarity, re-rankers)
• wikipedia fine-tuning and dataset creation
• ReFT fine-tuning
• multi-gpu fine-tuning
• orpo fine-tuning
• memorization fine-tuning
• mamba-fine-tune
• DPO fine-tuning
• chat-fine-tuning
• function-calling fine-tuning
• long-context fine-tuning
• quantization
• supervised-fine-tuning
• unsupervised-fine-tuning

ADVANCED Inference

Get access here

• Server setup, API setup and inference guide for open source models
• Inference for function calling
• Monte Carlo Inference Methods
• Private Data Redaction
• Structured Data Extraction with LLMs
• Verified RAG citations

ADVANCED Vision Fine-tuning

Get access here

• Dataset Preparation
• Fine-tuning Llava, IDEFICS2

ADVANCED Transcription and Voice Cloning

Get access here

• Dataset Generation
• Fine-tuning Whisper speech-to-text on custom datasets
• Voice Cloning Fine-tuning / Text-to-Speech

One-click LLM API/Server Templates

• Check out the full list here

Function-calling Models and Datasets

Get access here

• Fine-tuned models for function calling - from 1.3B to 70B in size, incl. Mixtral, Yi, Deepseek, Llama and more. Collection here
• Dataset to fine-tune models for function calling. Access here
• Scripts to fine-tune models for function calling and structured responses.