PlaidML
Pricing Details
Free and open-source. Distributed under the Apache 2.0 license.Features
Multi-hardware acceleration (CPU, GPU); Supports AMD, Intel, NVIDIA GPUs; Framework integrations (Keras, ONNX); Open-source; No proprietary library dependencies (e.g., CUDA); Automatic code optimization; Supports Linux, macOS, Windows; C/C++ and Python APIs; Flexible and extensible architecture.Integrations
Integration with Keras, ONNX; Python, C, C++ APIs; Support for OpenCL, LLVM; Integration with Bazel (for building).Preview
PlaidML is an open-source machine learning acceleration framework designed to enable efficient execution of deep learning models across a variety of hardware, including CPUs and GPUs from different vendors (AMD, Intel, NVIDIA). A key feature of PlaidML is its independence from proprietary libraries like NVIDIA's CUDA, making it more versatile and accessible for use on diverse hardware. The framework provides a backend for popular deep learning libraries such as Keras and ONNX, allowing developers to run their existing models with acceleration on supported hardware. PlaidML includes components for automatically optimizing code for the specific hardware architecture, which helps improve performance. It supports major operating systems: Linux, macOS, and Windows. PlaidML is an open-source project, contributing to its flexibility and extensibility. Although the last official release on PyPI is from early 2020, the project remains active on GitHub and can be used for deep learning tasks on hardware not fully supported by other frameworks.