Model Summary

This model is a fine-tuned version of LLaMA 3.2-1B, optimized using LoRA (Low-Rank Adaptation) on the NVIDIA ChatQA-Training-Data. It is tailored for conversational AI, question answering, and other instruction-following tasks, with support for sequences up to 1024 tokens.

---

Key Features

• Base Model: LLaMA 3.2-1B
• Fine-Tuning Framework: LoRA
• Dataset: NVIDIA ChatQA-Training-Data
• Max Sequence Length: 1024 tokens
• Use Case: Instruction-based tasks, question answering, conversational AI.

Model Usage

This fine-tuned model is suitable for:

• Conversational AI: Chatbots and dialogue agents with improved contextual understanding.
• Question Answering: Generating concise and accurate answers to user queries.
• Instruction Following: Responding to structured prompts.
• Long-Context Tasks: Processing sequences up to 1024 tokens for long-text reasoning.

How to Use DoeyLLM / OneLLM-Doey-V1-Llama-3.2-1B-Instruct

This guide explains how to use the DoeyLLM model on both app (iOS) and PC platforms.

---

App: Use with OneLLM

OneLLM brings versatile large language models (LLMs) to your device—Llama, Gemma, Qwen, Mistral, and more. Enjoy private, offline GPT and AI tools tailored to your needs.

With OneLLM, experience the capabilities of leading-edge language models directly on your device, all without an internet connection. Get fast, reliable, and intelligent responses, while keeping your data secure with local processing.

Quick Start for mobile

Follow these steps to integrate the DoeyLLM model using the OneLLM app:

1. Download OneLLM
   Get the app from the App Store and install it on your iOS device.

   Or get the app from the Play Store and install it on your Android device.

2. Load the DoeyLLM Model
   Use the OneLLM interface to load the DoeyLLM model directly into the app:

   • Navigate to the Model Library.
   • Search for DoeyLLM.
   • Select the model and tap Download to store it locally on your device.

3. Start Conversing
   Once the model is loaded, you can begin interacting with it through the app's chat interface. For example:

   • Tap the Chat tab.
   • Type your question or prompt, such as:

     "Explain the significance of AI in education."

   • Receive real-time, intelligent responses generated locally.

Key Features of OneLLM

• Versatile Models: Supports various LLMs, including Llama, Gemma, and Qwen.
• Private & Secure: All processing occurs locally on your device, ensuring data privacy.
• Offline Capability: Use the app without requiring an internet connection.
• Fast Performance: Optimized for mobile devices, delivering low-latency responses.

For more details or support, visit the OneLLM App Store page and Play Store.

PC: Use with Transformers

The DoeyLLM model can also be used on PC platforms through the transformers library, enabling robust and scalable inference for various NLP tasks.

Quick Start for PC

Follow these steps to use the model with Transformers:

1. Install Transformers
   Ensure you have transformers >= 4.43.0 installed. Update or install it via pip:

   pip install --upgrade transformers
   

2. Load the Model
   Use the transformers library to load the model and tokenizer:

Starting with transformers >= 4.43.0 onward, you can run conversational inference using the Transformers pipeline abstraction or by leveraging the Auto classes with the generate() function.

Make sure to update your transformers installation via pip install --upgrade transformers.

import torch
from transformers import pipeline

model_id = "OneLLM-Doey-V1-Llama-3.2-1B"
pipe = pipeline(
    "text-generation",
    model=model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
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])

Responsibility & Safety

As part of our responsible release strategy, we adopted a three-pronged approach to managing trust and safety risks:

Enable developers to deploy helpful, safe, and flexible experiences for their target audience and the use cases supported by the model. Protect developers from adversarial users attempting to exploit the model’s capabilities to potentially cause harm. Provide safeguards for the community to help prevent the misuse of the model.