Athene V2 Chat

Introduction
Athene-V2-Chat-72B is an advanced open-weights large language model (LLM) developed by Nexusflow. It rivals GPT-4o across various benchmarks, excelling in chat, math, and coding capabilities. It is known for its performance in the Chatbot Arena, surpassing GPT-4o-0513 in several categories.

Architecture
Athene-V2-Chat-72B is based on the Qwen-2.5-72B-Instruct model and is fine-tuned using Reinforcement Learning from Human Feedback (RLHF). This model is designed to handle conversational tasks efficiently, making it highly suitable for chat applications.

Training
The model is trained with a focus on improving its performance in chat, math, and coding tasks. Training involved using the Qwen-2.5-72B-Instruct as the base model, and further optimization was achieved through RLHF, enhancing its ability to follow instructions and handle multi-turn interactions.

Guide: Running Locally
To run Athene-V2-Chat-72B locally, you can use the Transformers library from Hugging Face. Here are the basic steps:

1. Install the Transformers library:

   pip install transformers
   

2. Load the model and tokenizer:

   from transformers import AutoModelForCausalLM, AutoTokenizer
   
   model_name = "Nexusflow/Athene-V2-Chat"
   model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype="auto", device_map="auto")
   tokenizer = AutoTokenizer.from_pretrained(model_name)
   

3. Prepare a prompt and generate a response:

   prompt = "Write a Python function to return the nth Fibonacci number in log n runtime."
   messages = [{"role": "user", "content": prompt}]
   text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
   model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
   generated_ids = model.generate(**model_inputs, max_new_tokens=2048)
   response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
   

For optimal performance, using a cloud GPU is recommended, especially for models with large parameters like Athene-V2-Chat-72B.

License
Athene-V2-Chat-72B is distributed under the Nexusflow Research License. You can view the full license here.