This project implements a character-level text generation model using PyTorch. The model is trained on text data and can generate new sequences of text based on a seed input.
- Custom Dataset: The project defines a dataset class that converts input text into sequences of characters, which are then used for training the model.
- LSTM-based Model: The model uses an LSTM (Long Short-Term Memory) architecture to learn from the input sequences and generate text.
- Text Generation: After training, the model can generate new text by predicting character sequences from a starting seed.
- GPU Support: Utilizes GPU if available for faster training.
The dataset used for training the model is a portion of the text from Pride and Prejudice by Jane Austen. The text is converted into sequences of characters for training.
To run this project, you need to install the following dependencies:
torch: For building and training the model.numpy: For array manipulations.
You can install the dependencies using pip:
pip install torch numpy- TextDataset: A PyTorch dataset that prepares the input text for training by converting it into sequences of character indices.
- TextGeneratorModel: The LSTM-based model that takes input sequences and generates character predictions.
- Training Function: A function that trains the model using cross-entropy loss and Adam optimizer.
- Text Generation Function: A function that generates text from a trained model by feeding it a starting sequence.
- Main Script: A script to load data, train the model, save the trained model, and generate text.
The model consists of:
Embedding Layer: Converts input character indices into dense vectors. LSTM Layer: Learns temporal dependencies from the input sequences. Fully Connected Layer: Outputs probabilities for the next character in the sequence.
To run this project, clone the repository and run the following command,
python main.pyLicense This project is licensed under the GNU License.