MISHANM/meta-Llama-3.3-70B-Instruct-fp16

This model is a FP16 quantized version of the meta-llama/Llama-3.3-70B-Instruct, optimized for efficient inference with reduced memory usage while maintaining high precision, supported languages : English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai.

Model Details

1. Tasks: Causal Language Modeling, Text Generation
2. Base Model: meta-llama/Llama-3.3-70B-Instruct
3. Quantization Format: FP16

Device Used

1. GPUs: AMD Instinct™ MI210 Accelerators

Inference with HuggingFace

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

# Load the fine-tuned model and tokenizer
model_path = "MISHANM/meta-Llama-3.3-70B-Instruct-fp16"

model = AutoModelForCausalLM.from_pretrained(model_path,device_map="auto")

tokenizer = AutoTokenizer.from_pretrained(model_path)

# Function to generate text
def generate_text(prompt, max_length=1000, temperature=0.9):
   # Format the prompt according to the chat template
   messages = [
       {
           "role": "system",
           "content": "Give response to the user query.", # change as per your requirement.
       },
       {"role": "user", "content": prompt}
   ]

   # Apply the chat template
   formatted_prompt = f"<|system|>{messages[0]['content']}<|user|>{messages[1]['content']}<|assistant|>"

   # Tokenize and generate output
   inputs = tokenizer(formatted_prompt, return_tensors="pt")
   output = model.generate(  # Use model.module for DataParallel
       **inputs, max_new_tokens=max_length, temperature=temperature, do_sample=True
   )
   return tokenizer.decode(output[0], skip_special_tokens=True)

# Example usage
prompt = """Give a poem on LLM ."""
text = generate_text(prompt)
print(text)

Citation Information

@misc{MISHANM/meta-Llama-3.3-70B-Instruct-fp16,
  author = {Mishan Maurya},
  title = {Introducing FP16 quantized version of meta-llama/Llama-3.3-70B-Instruct},
  year = {2024},
  publisher = {Hugging Face},
  journal = {Hugging Face repository},
  
}