This model is the foundational LLM engine for the Uruti AI-powered incubator. It has been explicitly fine-tuned on localized Rwandan tech ecosystem data, startup advisory frameworks, and domain-specific instruction sets.

Base Architecture: Qwen/Qwen2.5-7B-Instruct Precision: bfloat16 Merge Status: Fully baked (No separate LoRA adapter required).

Backend Integration (FastAPI / vLLM)

This directory contains the complete, merged weights. It can be loaded directly into any modern inference engine.

Using Hugging Face Transformers

python

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_path = "./uruti_production_model"
tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModelForCausalLM.from_pretrained(
    model_path,
    device_map="auto",
    torch_dtype=torch.bfloat16
)

# System Prompt Requirement
system_prompt = {"role": "system", "content": "You are an Uruti AI advisory assistant specialized in the Rwandan ecosystem."}

NiyonshutiDavid/uruti-advisory-model-best

Uruti AI Advisory Model (Production Ready)

Overview

Backend Integration (FastAPI / vLLM)

Using Hugging Face Transformers

NiyonshutiDavid/
uruti-advisory-model-best