Nicat-dcw/GreesyGuard

GreesyGuard

About

GreesyGuard is a text moderation model trained on Gitpod to identify and filter inappropriate content. GreesyGuard created with the help of claude-3.5-sonnet

Installation Steps

1. Clone the repository:

   git clone https://github.com/Nicat-dcw/greesyguard.git
   cd greesyguard

2. Install the required packages:

   pip install -r requirements.txt

3. Prepare your dataset:

   Ensure your dataset contains fields tweet and label.

4. Train the model:

   python train.py

5. Run inference:

   python inference.py

---

Benchmark

We used the dataset for training the model for benchmark

	HumanEval	SG Prompt
GreesyGuard-2	42%	89.7
Text-mod-007	16%	85.6
ShieldGemma(2b)	No Data	No Data

Changes in this version

• Increased Vocab size
• Tokenizer (p50>cl100k)
• Max length (128>2048)
• Learning rate (2e-5)
• Hugginface's datasets support
• Better learning handling
• API Support (OpenAI)

Next version: 10 stars