TinyGPT

TinyGPT is a lightweight and Fully customizable GPT-style network made from scratch.
Designed for developers who want to define the limits of the model freely, or for beginners who want to experiment with AI and learn by hands-on experience, TinyGPT makes it easy to experiment with GPT architectures without heavy infrastructure requirements.

Observation: TinyGPT is looking forward to updates and new versions.

---

Features

• Fully customizable GPT skeleton
• Specify model parameters and sizes like vocab_size n_layer, n_head, n_embd, and seq_length
• Custom dataset
• Simple Skeleton for adaptation to many enviroments

---

System minimum recommendations

• CPU: Dual-core or better; 2.0 GHz recommended for CPU-only runs.
• RAM: At least 6 GB.
• Disk Space: 2-20 GB free (HDD or SSD) depending on dataset size and frameworks, but for small projects, 2GB its just fine.
• GPU: Dedicated GPU with ≥1 GB VRAM recommended for PyTorch. CUDA requires proper GPU drivers.

PyTorch Drivers: Visit The official page to Install the CUDA Drivers: https://pytorch.org/get-started/locally/

---

Installation

# Clone the repository
git clone https://github.com/mnisperuza/TinyGPT.git

# Navigate into the repository
cd TinyGPT

# Install dependencies
pip install -r tinyGPT_(version)/requirements.txt