arberbr/bill-nye-science-guy

Bill Nye Science Guy

A fine-tuned version of DialoGPT Medium that responds in the style of Bill Nye, the Science Guy.

Model Description

This model has been fine-tuned using LoRA (Low-Rank Adaptation) to emulate Bill Nye's characteristic enthusiastic and educational communication style. It's designed to provide scientific explanations and fact-checking responses in a way that's engaging, accessible, and true to Bill Nye's personality.

Training Data

The model was trained on a synthetic dataset of 1,500 examples that include:

• General science Q&A responses in Bill Nye's style
• Fact-checking scenarios with scientific explanations
• Educational content covering various scientific topics

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch

# Load base model and tokenizer
base_model = "microsoft/DialoGPT-medium"
tokenizer = AutoTokenizer.from_pretrained(base_model)
model = AutoModelForCausalLM.from_pretrained(base_model)

# Load LoRA adapter
model = PeftModel.from_pretrained(model, "arberbr/bill-nye-science-guy")

# Generate response
prompt = "Human: What is climate change?\n\nBill Nye:"
inputs = tokenizer(prompt, return_tensors="pt")

with torch.no_grad():
    outputs = model.generate(**inputs, max_new_tokens=200, temperature=0.7)
    
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)

Training Details

• Base Model: microsoft/DialoGPT-medium
• Fine-tuning Method: LoRA (Low-Rank Adaptation)
• Training Examples: 1,500
• Epochs: 3
• Learning Rate: 2e-4
• Batch Size: 2
• LoRA Rank: 16
• LoRA Alpha: 32

Limitations

• This model is designed for educational and entertainment purposes
• Responses should be fact-checked for accuracy
• The model may not always provide scientifically accurate information
• Use with caution in professional or critical applications

License

This model is released under the Apache 2.0 license.