abdullah75f/Transformer-Amharic-bot-new-colab

This Google Colab notebook contains the complete process for fine-tuning rasyosef/gpt2-small-amharic. It handles data loading, cleaning, tokenization, training via Hugging Face, and model saving # Transformer-Amharic-Bot-New-Colab This repository details the fine-tuning of `rasyosef/gpt2-small-amharic` for improved Amharic text generation using Google Colab and Hugging Face Trainer. The notebook covers loading a custom corpus (from Google Drive), data cleaning, tokenization, training, and evaluation. Includes a command-line test chatbot within the notebook and setup instructions for a local Streamlit application. **✨ Live Demo:** **abdullah75f-transformer-amh… ✨ ## Features * Fine-tunes `rasyosef/gpt2-small-amharic` on a custom Amharic corpus. * Handles data cleaning, tokenization, training, and evaluation. * Saves the trained model to Google Drive. * Includes a basic command-line chatbot in the notebook. * Provides setup for a local Streamlit chatbot (requires `chatbot_app.py`). ## Screenshots ## Screenshots ### Chatbot User Interface (Streamlit) ### Model Explanations (Streamlit Sidebar) ### Model Internals Example (Streamlit Calculations) ## Setup and Usage ### 1. Google Colab (Training) 1. **Open Notebook:** Upload and open the `.ipynb` file in Google Colab. 2. **Data:** Place your `raw-corpus.txt` in the Google Drive path specified in the notebook config (e.g., `/content/drive/MyDrive/Amharic_Chatbot/`). 3. **Run:** Execute all cells (`Runtime` -> `Run all`). Connect Drive when prompted. 4. **Output:** The fine-tuned model is saved to Google Drive (e.g., `/content/drive/MyDrive/Amharic_Chatbot/models/amharic-gpt-finetuned/final-new`). ### 2. Local Chatbot (Streamlit - Requires `chatbot_app.py`) 1. **Clone Repo & Get Model:** Clone this repository and download the saved fine-tuned model files from your Google Drive. Place the model files in the path expected by `chatbot_app.py` (e.g., `./model/final-new/`). 2. **Setup Environment:** ```bash # Create & activate a virtual environment (recommended) python3 -m venv venv source venv/bin/activate # Adjust for your …