Noromaid 7 B 0.4 D P O

Introduction

Noromaid-7B-0.4-DPO is a collaborative text generation model developed by IkariDev and Undi. This model is optimized for running with FP16 precision and is designed to produce human-like responses using a specialized prompt format called CHATML.

Architecture

The model leverages the Mistral architecture, utilizing transformers to facilitate text generation. It supports inference through the text-generation-inference library and can be deployed using inference endpoints compatible with Hugging Face's ecosystem.

Training

The model was fine-tuned using a variety of datasets to enhance its output and human-like behavior. Key datasets include:

• No Robots Dataset: Enhances human-like behavior.
• Aesir Private RP Datasets: Fresh data from new datasets.
• LimaRP Dataset: Provides additional contextual data.

The model also underwent DPO (Data Protection Officer) training with datasets such as Intel/orca_dpo_pairs and NobodyExistsOnTheInternet/ToxicDPOqa to manage sensitive data responsibly.

Guide: Running Locally

To run this model locally:

1. Install Dependencies: Ensure you have Python and the necessary libraries installed, such as transformers and safetensors.
2. Clone the Repository: Download the model files from the Hugging Face model card.
3. Load the Model:

   from transformers import AutoModelForCausalLM, AutoTokenizer
   model = AutoModelForCausalLM.from_pretrained("NeverSleep/Noromaid-7B-0.4-DPO", torch_dtype=torch.float16)
   tokenizer = AutoTokenizer.from_pretrained("NeverSleep/Noromaid-7B-0.4-DPO")
   

4. Generate Text: Use the model for text generation by encoding your prompt and decoding the model's output.

For optimal performance, consider using cloud GPUs such as those offered by AWS, Google Cloud, or Azure.

License

The model is released under the CC BY-NC 4.0 license, allowing for non-commercial use with attribution.