Intern
    Data Science Chair

    Studying The Effectiveness Of Mamba

    10.02.2025

    The rise of ChatGPT has introduced a wide range of AI-powered systems. However, despite the continuous advancements, the underlying Transformer architecture still faces limitations. With this thesis, you will explore different architectures like Mamba and the effects of their core parts.

    The rise of ChatGPT has introduced a wide range of AI-powered services. However, despite continuous advancements, the underlying Transformer architecture still faces limitations which led to the exploration of Mamba.

    This thesis focuses on understanding how different components of Mamba contribute to its overall effectiveness. By isolating specific elements, we aim to identify which parts have the most significant impact on performance.

    Requirements:
    - Python
    - PyTorch
    Nice to have:
    - HuggingFace Transformers
    - HuggingFace PEFT

    Supervisor: Anton Vlasjuk

    Zurück