DanielFDias/Llama4B_fantasyStoriesUpdated

[Llama-Fantasy-Story-Generator]

Model Description

This model is a Large Language Model (LLM) based on the Llama architecture, specifically adapted [or fine-tuned] to generate short fantasy stories. It is designed to take a structured prompt containing keywords, characters, and plot points (beginning, midway, conclusion) and generate a narrative incorporating these elements.

This model was developed as part of a project documented in this Kaggle Notebook: https://www.kaggle.com/code/danieldias1998/isep-llm-fantasystories

Intended Uses & Limitations

Intended Use

This model is intended for:

• Generating creative fantasy story snippets based on user-provided structural elements.
• Experimentation with controlled text generation in the fantasy genre.
• Serving as a creative aid or inspiration for writers.

Limitations

• Prompt Dependency: The quality and coherence of the generated story heavily depend on the clarity and consistency of the input prompt.
• Strict Adherence: The model may not always perfectly adhere to every detail in the prompt structure, especially nuances in the Midway and Conclusion. It uses the prompt as strong guidance rather than absolute constraints.
• Repetition & Coherence: Like many LLMs, it may sometimes generate repetitive text or lose coherence over longer passages (though max_length should be controlled).
• Factuality & Bias: The model generates fictional content. It may inherit biases present in the base Llama model or any fine-tuning data. It should not be used for factual information.
• Context Length: The effective context length might be limited [mention if you know the specific limit].

How to Use

This model expects a prompt formatted exactly as follows. Ensure each section (Keywords, Characters, Beginning, Midway, Conclusion) starts on a new line.

from transformers import pipeline

# Replace '[your-huggingface-username/your-model-name]' with your actual model ID on the Hub
generator = pipeline('text-generation', model='[your-huggingface-username/your-model-name]')

prompt = """Keywords: castle, dragon, sword
Characters: A dragon named Ignis, a brave knight Sir Kaelan
Beginning: Sir Kaelan heard tales of a fierce dragon, Ignis, guarding an ancient castle and decided to investigate.
Midway: Upon meeting Ignis, Sir Kaelan discovered the dragon was misunderstood and lonely. They formed an unlikely friendship, exploring the castle's secrets together.
Conclusion: When a shadow army threatened the kingdom, Sir Kaelan, riding Ignis, led the charge, saving the realm and proving friendship is stronger than fear."""

# It's recommended to specify max_length and potentially other parameters
# Adjust max_length based on expected story size
generated_stories = generator(
    prompt,
    max_length=500, # Adjust as needed
    num_return_sequences=1,
    # Add other parameters like do_sample=True, temperature=0.7, top_k=50 if desired
)

for story in generated_stories:
    print(story['generated_text'])

# The output will start with the prompt, followed by the generated story.
# You might need to process the output string to separate the prompt from the generation.