unifyai/ivy
Ivy
Ivy is a machine learning framework transpiler and model converter designed to ensure deep learning portability. It serves as a tool for migrating source code and models between different deep learning frameworks while maintaining original functionality.
The system enables cross-framework model portability by translating model weights, architectures, and source code. It uses abstract syntax tree based transpilation and computational graph tracing to capture execution flows and rewrite high-level logic into target framework code.
The project covers model interoperability through weight-layout remapping and intermediate representation mapping. These capabilities allow for the extraction of parameter tensors from a source model and their conversion into the memory layout of a target framework.
Features
- Backend-Agnostic Deep Learning - Provides a backend-agnostic deep learning environment by decoupling mathematical logic from specific library APIs.
- Computational Graphs - Uses computational graphs to represent mathematical operations for framework-independent analysis and execution.
- Cross-Framework Model Conversion - Translates model weights and architectures between different machine learning frameworks for target-specific execution.
- Deep Learning Portability Tools - Maintains neural network functionality when migrating code and parameters between different machine learning environments.
- Framework Migration Tools - Facilitates moving model code and libraries between deep learning frameworks to avoid manual rewriting.
- Machine Learning Transpilers - Translates model definitions and functional logic between different deep learning frameworks.
- Machine Learning Model Portability - Ensures deep learning models can execute across different platforms by translating their underlying structure.
- Model Migrators - Provides tools for migrating model weights and definitions between different deep learning frameworks.
- Model Format Converters - Translates model weights and architectures between different framework formats for compatible execution.
- Framework Weight Remappers - Transforms parameter tensors from a source model to match the memory layout of a target framework.
- Source-To-Source Transpilers - Implements source-to-source transpilation to convert machine learning code between different framework syntaxes.
- Computational Graph Tracers - Captures the execution flow of functions as directed graphs to enable framework-independent analysis.
- Intermediate Representation Translation - Translates framework-specific operations into a common intermediate representation to facilitate cross-platform conversion.
- Operator Dispatchers - Routes computational operations to the appropriate backend runtime based on the target framework's available functions.
- Graph Execution Tracing - Captures execution flows as graphs to analyze or execute operations within a target framework.
- Deep Learning Frameworks - Unified ML framework transpiler supporting multiple backend libraries.
Star history
How this analysis was created: This summary and feature list are AI-generated from collected project material and can contain mistakes. Stars, license and language are imported from GitHub. Inclusion does not mean that we have tested or audited this project. Check the source documentation for any feature you depend on. Learn more on our About page.
