GOAT-AI/GOAT-70B-Storytelling
GOAT-AI/GOAT-70B-Storytelling is a 69 billion parameter LLaMA 2-based causal language model developed by GOAT.AI lab, specifically fine-tuned for generating high-quality, cohesive narratives. With a context window of 4096 tokens, it is optimized for creative writing tasks such as producing books, novels, and movie scripts. This model is designed to function as a core component within an autonomous story-writing agent.
Loading preview...
GOAT-70B-Storytelling: Narrative Generation Model
GOAT-70B-Storytelling is a 69 billion parameter model from GOAT.AI lab, built upon the LLaMA 2 architecture and licensed under LLaMA-2. It is primarily designed as the core engine for an autonomous story-writing agent, focusing on generating extensive and cohesive narratives.
Key Capabilities & Features
- Specialized Narrative Generation: Optimized for creating high-quality, cohesive, and captivating stories, novels, and movie scripts.
- Agent Integration: Intended for use within the GOAT-Storytelling-Agent, which processes plot outlines, character profiles, and relationships to produce detailed narratives.
- Training Details: Instruction fine-tuned on 18,000 examples over one epoch using a cluster of 64xH100 GPUs, employing FSDP ZeRO-3 sharding.
- Context Window: Features a 4096-token context window, suitable for managing narrative flow.
Performance & Limitations
Evaluated on the Open LLM Leaderboard, the model achieved an average score of 67.38, with specific scores including 68.77 on AI2 Reasoning Challenge and 69.92 on MMLU. Users should be aware that, like other large language models, GOAT-70B-Storytelling can produce factually incorrect, biased, or otherwise offensive outputs and should not be relied upon for factual accuracy.
When to Use This Model
This model is ideal for developers and storytellers looking to automate or assist in the creation of long-form written content, particularly for generating fictional narratives. Its design as an agent component makes it suitable for complex story development workflows.