Local LLM Training Studio

A lightweight software to train, finetune, and test a small Language Model (LLM) on your local CPU. No heavy GPUs required!

Features

• Custom Data Training: Input your own USER INPUT / OUTPUT pairs.
• Adjustable Hyperparameters: Control Learning Rate, Batch Size, and Epochs.
• Inference Playground: Test your model with adjustable Temperature and Nucleus Sampling (Top-P).
• Lightweight: Uses a miniature GPT architecture implemented in pure PyTorch (no Hugging Face transformers dependency).
• CPU Friendly: Designed to run on standard laptops.

Prerequisites

• Python 3.8 or higher

Installation

1. Open a terminal in this folder.

2. Install the required packages:

   pip install -r requirements.txt

Usage

1. Run the application:

   streamlit run app.py

2. Tab 1 (Data):

   • Paste your training data in the text box.
   • Click Parse & Preview to prepare the data.

3. Tab 2 (Training):

   • Set your desired parameters.
   • Click Start Training. Watch the loss go down!

4. Tab 3 (Inference):

   • Once training is complete, switch to this tab.
   • Type a message and see how your model responds.
   • Play with Temperature and Top-P to change the creativity.

How it works

This software implements a NanoGPT (a small Transformer-based language model) from scratch. It treats your input data as a continuous stream of text and learns to predict the next character. It uses a custom character-level tokenizer for simplicity.

File Structure

• app.py: The main Streamlit interface.
• model.py: The GPT model architecture (PyTorch).
• train.py: Training loop and logic.
• utils.py: Data parsing and tokenization utilities.