llama-3.2-3b-support-ticket-json

llama-3.2-3b-support-ticket-json

A LoRA fine-tune of meta-llama/Llama-3.2-3B-Instruct that converts free-text customer support messages into structured JSON with category, priority, and summary fields.

Trained on 35 synthetic support-ticket examples using TRL's SFTTrainer with LoRA (r=16, alpha=32, targeting the attention projection layers). Built with Llama.

Quick start

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

base_model = "meta-llama/Llama-3.2-3B-Instruct"
adapter = "Kranthi1290/llama-3.2-3b-support-ticket-json"

tok = AutoTokenizer.from_pretrained(base_model)
model = AutoModelForCausalLM.from_pretrained(base_model, torch_dtype=torch.bfloat16, device_map="auto")
model = PeftModel.from_pretrained(model, adapter)

messages = [
    {"role": "system", "content": "You extract structured info from customer support messages. Respond with ONLY a JSON object with keys: category, priority, summary. No other text."},
    {"role": "user", "content": "My login keeps failing after the update, this is urgent, please help asap"},
]
prompt = tok.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tok(prompt, return_tensors="pt").to(model.device)
out = model.generate(**inputs, max_new_tokens=100, do_sample=False)
print(tok.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))

Framework versions

• PEFT 0.19.1
• TRL: 1.8.0
• Transformers: 5.14.1
• Pytorch: 2.10.0a0+b558c986e8.nv25.11
• Datasets: 5.0.0
• Tokenizers: 0.22.1