attentionmech/mav
A command-line tool for visualizing attention patterns, token activations, and internal mechanics of large language models during text generation.
Not currently ranked — collecting fresh signals.
star history
MAV (Model Activity Visualiser) provides real-time visualization of LLM internals including attention patterns, token probabilities, and layer activations. It runs as a CLI application and integrates with Hugging Face transformers models, allowing developers and researchers to inspect model behavior during inference. The tool supports multiple open-source models like GPT-2, SmolLM, and Llama, displaying activation data in an interactive visual format to aid interpretability research and debugging.
Frequently asked
- What is attentionmech/mav?
- A command-line tool for visualizing attention patterns, token activations, and internal mechanics of large language models during text generation.
- Is mav open source?
- Yes — attentionmech/mav is open source, released under the MIT license.
- What language is mav written in?
- attentionmech/mav is primarily written in Python.
- How popular is mav?
- attentionmech/mav has 523 stars on GitHub.
- Where can I find mav?
- attentionmech/mav is on GitHub at https://github.com/attentionmech/mav.