IT Helpdesk AI Classifier — v4
Fine-tuned LLM for classifying IT helpdesk tickets into categories, subcategories, and generating insights for corporate IT support teams.
Model Details
- Base Model: mdk615661/it-helpdesk-merged-v3
- Fine-tuning: QLoRA (LoRA r=16, alpha=32)
- Training Data: 2,000 IT helpdesk records (Qwen-generated)
- Training Loss: 0.187 (3 epochs)
- Hardware: Kaggle T4 GPU (33 min)
- Precision: fp16
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
model_name = "mdk615661/it-helpdesk-merged-v4"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
model_name,
dtype=torch.float16,
device_map="auto"
)
prompt = """### Instruction:
Normalize and classify this IT helpdesk ticket.
### Input:
Laptop is not turning on
### Output:
"""
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
outputs = model.generate(
**inputs,
max_new_tokens=150,
do_sample=False,
repetition_penalty=1.3,
pad_token_id=tokenizer.eos_token_id
)
print(tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True))
Output Format
Category: Hardware
SubCategory: Hardware - Laptop
Normalized: laptop not working
Priority: Medium
Insight: Hardware failure preventing user from working.
Recommendation: Raise repair request with IT hardware team.
Categories
| Category |
Subcategories |
| Hardware |
Laptop, Charger, Mobile |
| Software |
Installation, VPN, Password Reset, O365, Teams Issue, MFA Reset, BitLocker, OS Installation, USB Access |
| Incident |
Critical Incident, Network Outage, Security Incident, Service Outage, Performance Incident |
| Procurement |
Hardware Procurement, Software Procurement |
| Onboarding & Offboarding |
Onboarding, Offboarding |
| Cloud & Infrastructure |
DR and BCP, Infrastructure Configuration, Network Configuration Request, System Configuration |
| Asset |
Asset Management, Asset Request, Asset - Client |
| Others |
Account Management, Audit and Compliance, Change Management, IT Training Request, Vendor Support |
Training Hyperparameters
| Parameter |
Value |
| Epochs |
3 |
| Batch size |
4 |
| Gradient accumulation |
4 |
| Learning rate |
2e-4 |
| Warmup steps |
50 |
| LR scheduler |
cosine |
| LoRA r |
16 |
| LoRA alpha |
32 |
Version History
| Version |
Training Data |
Final Loss |
| v3 |
1,141 real TruMIS tickets |
— |
| v4 (this) |
+ 2,000 Qwen records |
0.187 |
