This is the Linux app named Penzai whose latest release can be downloaded as Penzai0.2.5sourcecode.tar.gz. It can be run online in the free hosting provider OnWorks for workstations.
Download and run online this app named Penzai with OnWorks for free.
Дотримуйтесь цих інструкцій, щоб запустити цю програму:
- 1. Завантажив цю програму на свій ПК.
- 2. Введіть у наш файловий менеджер https://www.onworks.net/myfiles.php?username=XXXXX із потрібним ім'ям користувача.
- 3. Завантажте цю програму в такий файловий менеджер.
- 4. Запустіть онлайн-емулятор OnWorks Linux або Windows або онлайн-емулятор MACOS з цього веб-сайту.
- 5. З ОС OnWorks Linux, яку ви щойно запустили, перейдіть до нашого файлового менеджера https://www.onworks.net/myfiles.php?username=XXXXX з потрібним іменем користувача.
- 6. Завантажте програму, встановіть її та запустіть.
ЕКРАНИ
Ad
Penzai
ОПИС
Penzai, developed by Google DeepMind, is a JAX-based library for representing, visualizing, and manipulating neural network models as functional pytree data structures. It is designed to make machine learning research more interpretable and interactive, particularly for tasks like model surgery, ablation studies, architecture debugging, and interpretability research. Unlike conventional neural network libraries, Penzai exposes the full internal structure of models, enabling fine-grained inspection and modification after training. Its modular design includes tools for tree manipulation, named axes, and declarative neural network construction. The library integrates tightly with Treescope, an advanced pretty-printer for visualizing deeply nested JAX pytrees and NDArray structures. Penzai’s penzai.nn module provides a compositional, combinator-based API for building neural networks.
Функції
- Builds and represents models as readable JAX pytrees for full structural transparency
- Provides Treescope for interactive visualization of complex model hierarchies
- Includes pytree manipulation utilities for selective editing and rewrites (pz.select)
- Supports named axes for flexible vectorization and clear computation graphs (pz.nx)
- Combinator-based neural network library (pz.nn) with state and parameter sharing
- Modular transformer implementations for Gemma, LLaMA, Mistral, and GPT-NeoX architectures
Мова програмування
Python
Категорії
This is an application that can also be fetched from https://sourceforge.net/projects/penzai.mirror/. It has been hosted in OnWorks in order to be run online in an easiest way from one of our free Operative Systems.
