Opus V1 34b

Table of Contents

Model Overview

The DreamGen Opus V1 model is a cutting-edge AI designed for steerable story-writing and role-playing. It’s capable of generating engaging stories, summarizing plots, and even assisting with writing-related tasks.

Tasks and rank

Text Generation 16th

Natural Language Understanding 15th

Dialogue Systems 9th

Learn More About the Ranking System

Capabilities

Story-Writing and Role-Playing

The model uses an extended version of ChatML, allowing for interactive storytelling and role-playing. It can understand and respond to prompts, and even take on different character roles.

Writing-Related Tasks

The model can also perform various writing-related tasks, such as:

• Story Plot Summarization: Summarize a story or chapter in 50-200 words.
• Story Character Description: Describe a character from a story in 50 words.
• Story Style Description: Describe the style of a story, including atmosphere, setting, and literary devices.
• Story Description to Chapters: Generate a description for each chapter of a story based on a brief plot description and desired number of chapters.

Examples

Summarize the following short fiction story in 100 words: Malygris, a powerful magician, summons his long-dead lover Nylissa from the grave using necromancy. However, upon seeing her, he realizes that she is not the same person he once knew, leading him to question the nature of love, memory, and the power of magic. Malygris, a magician, uses necromancy to summon his deceased lover Nylissa. However, upon seeing her, he realizes she has changed, prompting him to question the nature of love, memory, and magic. He becomes disillusioned and dismisses her spirit, returning to his lonely tower as a weary sorcerer.

Describe the character of Malygris from the story in 50 words. Malygris is an ancient and powerful magician with a long white beard, cold green eyes, and a violet robe. He is a master of necromancy and possesses vast knowledge of forbidden magic. Despite his power, he is plagued by melancholy and ennui.

Write a continuation of the story from the perspective of Malygris. Describe his feelings after dismissing Nylissa's spirit. As I stood alone in my tower, the silence was deafening. I felt a hollow emptiness within me, a sense of loss and longing that I could not shake. I realized that even my mastery of necromancy could not bring back the past, and that true love was lost forever.

How it Works

To use the model, you’ll need to format your prompt correctly using the Opus V1 extension. This involves adding specific tags, like <|im_start|> and <|im_end|>, to indicate the start and end of each message.

Performance

The model is a powerhouse when it comes to performance. With 7B parameters, it can handle massive amounts of data quickly and efficiently. Whether you’re generating stories, role-playing, or summarizing articles, the model can keep up with your demands.

Speed

The model can process information quickly, making it ideal for tasks that require fast turnaround times.

Accuracy

The model delivers accurate results, thanks to its fine-tuning dataset of ~100M tokens.

Efficiency

The model is designed to be compatible with a range of software and hardware, making it easy to integrate into your workflow.

Alternatives

NeMo SFT with Llama 2 10500 Fine-tuned language model

Bert Base Uncased 1831 Uncased BERT model

Roberta Base 399 Masked language model

Limitations

While the model is powerful, it’s not perfect. Here are some limitations to keep in mind:

• Lack of consistency: The model may struggle to maintain consistency in the story or role-play, especially if the input is complex or open-ended.
• Limited contextual understanding: The model may not always grasp the nuances of human communication, leading to misunderstandings or misinterpretations.
• Over-reliance on patterns: The model may rely too heavily on patterns and structures learned from the training data, which can result in predictable and less engaging outputs.

Format

The model uses an extended version of ChatML, which requires specific formatting for input and output.

Input Format

To use the model, you’ll need to format your input in a specific way. Here’s an example:

<|im_start|>system
(Story description in the right format here)
(Typically consists of plot description, style description and characters)
<|im_end|>
<|im_start|>user
(Your instruction on how the story should continue)
<|im_end|>
<|im_start|>text names= Alice
(Continuation of the story from the Alice character)
<|im_end|>
<|im_start|>text
(Continuation of the story from no character in particular (pure narration))
<|im_end|>

Note the use of <|im_start|> and <|im_end|> to indicate the start and end of each message.

Running the Model

You can run the model on various platforms, including:

• DreamGen.com (free)
• Google Colab
• Locally, using a script or API

Make sure to read the prompt guide and formatting code to ensure correct tokenization and prompt formatting.