O L Mo 2 1124 13 B Instruct

Introduction

OLMo-2-1124-13B-Instruct is a powerful language model developed by AllenAI. It is built to generate human-like text and is primarily used for text generation tasks. The model is finetuned on diverse datasets for improved performance in various applications, including mathematical reasoning and conversational AI.

Architecture

OLMo-2-1124-13B-Instruct is the result of extensive training processes, including supervised fine-tuning, DPO training, and RLVR training. The model is based on the OLMo-2 architecture and uses a 13 billion parameter configuration. It is primarily trained in English and leverages the Tülu 3 dataset for state-of-the-art performance across numerous tasks.

Training

The model undergoes a comprehensive training pipeline that includes:

• Supervised Fine-tuning on a variant of the Tülu 3 dataset.
• DPO Training for further refinement on the same dataset.
• RLVR Training to enhance its ability to generate contextually relevant responses.

Guide: Running Locally

1. Installation: Ensure you have the latest version of the Transformers library by installing it from the main branch:

   pip install --upgrade git+https://github.com/huggingface/transformers.git
   

2. Loading the Model: Use the following code snippet to load the model:

   from transformers import AutoModelForCausalLM
   
   olmo_model = AutoModelForCausalLM.from_pretrained("allenai/OLMo-2-1124-13B-Instruct")
   

3. Hardware Suggestions: Due to its size and complexity, it is recommended to use cloud GPUs for running the model efficiently. Services like AWS, Google Cloud, or Azure offer suitable GPU instances.

License

OLMo-2-1124-13B-Instruct is licensed under the Apache 2.0 license. It is intended for research and educational purposes. Users should refer to the Responsible Use Guidelines provided by AllenAI and adhere to additional terms associated with third-party datasets used in fine-tuning.