ChengManYu/qwen2-1.5b-ppt-outline-merged

Qwen2-1.5B PPT Outline Generator (Merged)

My first complete Hugging Face project 🎉

This is an AI model specifically designed to help students and professionals quickly generate professional 5-slide outlines.

Enter any topic (supports both Chinese and English), and you can generate a well-structured PPT outline suitable for direct use within seconds.

Model Features

• Foundation model：Qwen2-1.5B-Instruct
• Fine-tuning technology：LoRA + 4-bit 量化
• Training data：500 high-quality "Theme → PPT Outline" pairs
• Training platform：Kaggle GPU T4

Usage Method (Simplest)

from transformers import pipeline

pipe = pipeline(
    "text-generation", 
    model="ChengManYu/qwen2-1.5b-ppt-outline-merged",
    device_map="auto"
)

prompt = "Generate PPT outline for: AI in healthcare\n### Outline:"
output = pipe(prompt, max_new_tokens=400, temperature=0.7)[0]["generated_text"]
print(output.split("### Outline:")[-1])

Example

Input：Sustainable fashion Output：

• Slide 1: The Imperative for Sustainable Fashion
• Slide 2: Key Technologies and Practices
• Slide 3: Benefits and Challenges
• Slide 4: Real-World Success Stories
• Slide 5: The Path Forward

Training results

• Epochs: 3
• Final Validation Loss: 1.499
• The model converges stably and produces good quality outputs.

Restriction

• 1.5B parameter model, suitable for quick generation, but extremely complex topics may require multiple adjustments to the prompt
• Currently, the optimization for English prompts is the best. Chinese prompts are also supported, but their effectiveness is slightly weaker.