This project focuses on training a language model to predict the next token in a sequence using Long Short-Term Memory (LSTM) networks, implemented with PyTorch. The training data is sourced from a blue_caste.txt file, which provides the text corpus for learning. By leveraging LSTMs, the model maintains context and memory of the previous tokens, enabling it to generate coherent and contextually accurate text. This approach ensures that the generated sequences are more meaningful and aligned with natural language patterns. The model's architecture includes an embedding layer, multiple LSTM layers, and a fully connected layer that outputs the predicted token. The primary goal of this project is to enhance text generation capabilities by improving the model's ability to understand and predict subsequent tokens based on preceding ones. This project not only demonstrates the effective use of LSTMs for sequence prediction but also highlights the potential of PyTorch for building and training sophisticated neural network models.