titlebreaker-lora-adapter

Titlebreaker LoRA Adapter

This is a LoRA (Low-Rank Adaptation) adapter for the Qwen3-0.6B model, fine-tuned for title cleaning tasks.

Usage

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

# Load base model
base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-0.6B")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-0.6B")

# Load and apply the LoRA adapter
model = PeftModel.from_pretrained(base_model, "sch-ai/titlebreaker-lora-adapter")

# Generate clean title
def clean_title(dirty_title, max_length=200):
    prompt = f"<title_clean> "
    inputs = tokenizer(prompt, return_tensors="pt")
    
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_length=max_length,
            do_sample=True,
            temperature=0.7,
            pad_token_id=tokenizer.eos_token_id
        )
    
    generated = tokenizer.decode(outputs[0], skip_special_tokens=True)
    # Extract the clean title from between the tags
    if "</title_clean>" in generated:
        clean_title = generated.split("</title_clean>")[0].split("<title_clean>")[-1].strip()
        return clean_title
    return generated

# Example usage
dirty_title = "Your dirty title here"
clean_result = clean_title(dirty_title)
print(f"Clean title: {clean_result}")

Training Details

• Base model: Qwen/Qwen3-0.6B
• LoRA rank: 64
• LoRA alpha: 16
• LoRA dropout: 0.1
• Task type: Causal Language Modeling

Framework versions

• PEFT 0.17.0