Aurora-Fusion

Aurora-Fusion Model Card

Aurora-Fusion is a high-performance multilingual causal language model created by merging five state-of-the-art base models. It is optimized for instruction following, reasoning, and chat, and supports English, Spanish, and French natively, while maintaining multilingual capabilities.

Supported Base Models

• google/gemma-2-27b-it
• meta-llama/Llama-3-70B-Instruct
• mistralai/Mixtral-8x22B-Instruct-v0.1
• eleutherai/gpt-neox-20b

Training Datasets

• OpenHermes-2.5
• UltraChat-200k
• Open-Platypus
• MetaMathQA
• Wikipedia (multilingual)

Usage Example

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

tokenizer = AutoTokenizer.from_pretrained("YourUsername/Aurora-Fusion")
model = AutoModelForCausalLM.from_pretrained(
    "YourUsername/Aurora-Fusion",
    device_map="auto",
    load_in_4bit=True  # Optional: requires bitsandbytes
)

prompt = "Explain why Axolotls are so cute?"
inputs = tokenizer(prompt, return_tensors="pt")

outputs = model.generate(
    **inputs,
    max_new_tokens=250,
    do_sample=True,
    temperature=0.7,
    top_p=0.9
)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))