Triangulum 10 B

Introduction

Triangulum 10B is a series of large language models (LLMs) designed for multilingual applications. These models are pretrained and instruction-tuned for generative tasks, excelling in complex reasoning through long chains of thought.

Architecture

Triangulum 10B is built upon LLaMA's autoregressive language model, employing an optimized transformer architecture to enhance performance. It supports multiple languages, ensuring its utility across diverse linguistic contexts.

Training

The model's training involves:

1. Synthetic Datasets: Utilizes long chain-of-thought synthetic data for improved reasoning.
2. Supervised Fine-Tuning (SFT): Aligns the model with specific tasks using curated datasets.
3. Reinforcement Learning with Human Feedback (RLHF): Ensures alignment with human values and safety through iterative processes.

Guide: Running Locally

Step 1: Install Necessary Tools

Ensure you have the latest transformers library by running:

pip install --upgrade transformers

Step 2: Download the Model

Download the Triangulum-10B model using:

ollama run triangulum-10b-f16.gguf

Step 3: Interact with the Model

Once downloaded, you can interact with the model:

import torch
from transformers import pipeline

model_id = "prithivMLmods/Triangulum-10B"
pipe = pipeline(
    "text-generation",
    model=model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
messages = [
    {"role": "system", "content": "You are the kind and tri-intelligent assistant helping people to understand complex concepts."},
    {"role": "user", "content": "Who are you?"},
]
outputs = pipe(
    messages,
    max_new_tokens=256,
)
print(outputs[0]["generated_text"][-1])

Suggested Cloud GPUs

For better performance, consider using cloud GPUs from providers like AWS, Google Cloud, or Azure.

License

Triangulum 10B is distributed under the CreativeML OpenRAIL-M license, which allows for both academic and commercial use while maintaining ethical guidelines.