dynet
« Back to VersTracker
Description:
Dynamic Neural Network Toolkit
Type: Formula  |  Latest Version: 2.1.2@1  |  Tracked Since: Dec 17, 2025
Links: Homepage  |  formulae.brew.sh
Category: Ai ml
Tags: machine-learning deep-learning neural-networks nlp research
Install: brew install dynet
About:
DyNet is a neural network library designed for efficient implementation of models with dynamic structures, such as trees and graphs. Unlike frameworks with static computation graphs, it constructs graphs on-the-fly, offering flexibility for complex NLP and research tasks. It provides a C++ backend with bindings for Python and other languages.
Key Features:
  • Dynamic graph construction
  • High-performance C++ backend
  • Multi-language support (Python, C++)
  • Efficient batching for dynamic structures
Use Cases:
  • Natural Language Processing (NLP) research
  • Graph neural network implementation
  • Rapid prototyping of complex model architectures
Alternatives:
  • PyTorch – More widely adopted with a larger ecosystem, though DyNet is often cited for specific efficiency in certain NLP tasks.
  • TensorFlow – Uses a static graph approach by default (though Eager mode changes this), generally heavier on resource usage.
License: Apache-2.0
Dependencies: eigen@3
Bottles available for: arm64_tahoe, arm64_sequoia, arm64_sonoma, sonoma, arm64_linux, x86_64_linux
Version History
Detected Version Rev Change Commit
Sep 11, 2025 4:48am 0 VERSION_BUMP 775afd68
Sep 12, 2024 8:37am 0 VERSION_BUMP 833ae545