Llama Deepsync 1 B

Introduction

The Llama-Deepsync-1B is a fine-tuned version of the Llama-3.2-1B-Instruct model developed for text generation tasks. It is optimized for deep reasoning, logical structuring, and problem-solving, making it suitable for education, programming, and creative writing. It supports multilingual text generation and can generate step-by-step solutions and structured outputs like JSON.

Architecture

Llama 3.2 utilizes an auto-regressive language model with an optimized transformer architecture. It employs supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for helpfulness and safety. The model supports long-context capabilities, handling up to 128K tokens.

Training

The model is trained using a combination of supervised fine-tuning and reinforcement learning with human feedback to improve its performance in instruction following, coding, mathematics, and generating long texts. It also offers multilingual support for over 29 languages.

Guide: Running Locally

To run the Llama-Deepsync-1B model locally:

1. Install Transformers: Ensure you have transformers >= 4.43.0 by running pip install --upgrade transformers.
2. Load the Model: Use the Transformers library to load and run the model.

   import torch
   from transformers import pipeline
   
   model_id = "prithivMLmods/Llama-Deepsync-1B"
   pipe = pipeline(
       "text-generation",
       model=model_id,
       torch_dtype=torch.bfloat16,
       device_map="auto",
   )
   

3. Interact with the Model: Use the pipeline to generate text with role-based messages.

   messages = [
       {"role": "system", "content": "You are a pirate chatbot who always responds in pirate speak!"},
       {"role": "user", "content": "Who are you?"},
   ]
   outputs = pipe(
       messages,
       max_new_tokens=256,
   )
   print(outputs[0]["generated_text"][-1])
   

For cloud GPU support, consider platforms like AWS, Google Cloud, or Azure to enhance performance.

License

The Llama-Deepsync-1B model is licensed under the CreativeML Open RAIL-M license.