---
language: en
license: apache-2.0
tags:
  - legal
  - contract-analysis
  - clause-classification
  - onnx
  - quantized
  - int8
  - legal-bert
  - cuad
pipeline_tag: text-classification
base_model: nlpaueb/legal-bert-base-uncased
datasets:
  - theatticusproject/cuad-qa
library_name: optimum
---

# 🛡️ ClauseGuard CUAD Classifier — ONNX INT8 Quantized

**ONNX Runtime + dynamic INT8 quantized** version of the ClauseGuard legal clause classifier for **2-4x faster CPU inference**.

## Model Details

| Property | Value |
|----------|-------|
| Base model | `nlpaueb/legal-bert-base-uncased` (110M params) |
| LoRA adapter | `Mokshith31/legalbert-contract-clause-classification` |
| Export method | merge_and_unload() → ONNX → dynamic INT8 quantization |
| Labels | 41 CUAD clause categories |
| ONNX FP32 size | 438.3 MB |
| **INT8 size** | **110.3 MB (75% reduction)** |
| Runtime | ONNX Runtime (CPU) |
| Expected speedup | **2-4x** over PyTorch on CPU |

## Usage

```python
from optimum.onnxruntime import ORTModelForSequenceClassification
from transformers import AutoTokenizer
import torch

model = ORTModelForSequenceClassification.from_pretrained(
    "gaurv007/clauseguard-onnx-int8",
    file_name="model_quantized.onnx"
)
tokenizer = AutoTokenizer.from_pretrained("gaurv007/clauseguard-onnx-int8")

texts = ["The company may terminate your account at any time without notice."]
inputs = tokenizer(texts, return_tensors="pt", padding=True, truncation=True, max_length=512)

with torch.no_grad():
    outputs = model(**inputs)
    probs = torch.softmax(outputs.logits, dim=-1)
    predicted = probs.argmax(dim=-1)
    print(f"Label: {model.config.id2label[str(predicted.item())]}")
    print(f"Confidence: {probs.max().item():.3f}")
```

## 41 CUAD Labels

Document Name, Parties, Agreement Date, Effective Date, Expiration Date, Renewal Term, Notice Period to Terminate Renewal, Governing Law, Most Favored Nation, Non-Compete, Exclusivity, No-Solicit of Customers, No-Solicit of Employees, Non-Disparagement, Termination for Convenience, ROFR/ROFO/ROFN, Change of Control, Anti-Assignment, Revenue/Profit Sharing, Price Restriction, Minimum Commitment, Volume Restriction, IP Ownership Assignment, Joint IP Ownership, License Grant, Non-Transferable License, Affiliate License-Licensor, Affiliate License-Licensee, Unlimited/All-You-Can-Eat License, Irrevocable or Perpetual License, Source Code Escrow, Post-Termination Services, Audit Rights, Uncapped Liability, Cap on Liability, Liquidated Damages, Warranty Duration, Insurance, Covenant Not to Sue, Third Party Beneficiary, Other

## Part of ClauseGuard

This model powers [ClauseGuard](https://huggingface.co/spaces/gaurv007/ClauseGuard), the open-source legal contract analysis tool.

## License

Apache 2.0