plaidml/plaidml

Health v2 · maintenance only

Flags: archived

Adoption not part of the score

4566 stars · 398 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

PlaidML is a portable tensor compiler that enables deep learning on hardware (especially GPUs and embedded devices) not well supported by mainstream stacks. It sits underneath frameworks like Keras, ONNX, and nGraph, compiling tensor operations via an MLIR-based compiler stack.

Use cases

• run deep learning training on AMD GPUs without CUDA
• train Keras models on unsupported hardware
• compile tensor operations for embedded devices
• use deep learning on laptops without vendor-specific ML libraries
• avoid restrictive licensing in GPU ML software stacks

When to choose

• your GPU or device lacks support from TensorFlow or PyTorch
• you need a portable backend for Keras or ONNX models
• you want an open, permissively licensed tensor compiler

When to avoid

• you need cutting-edge performance on NVIDIA hardware with CUDA available
• you require actively maintained software - the master branch is frozen and v1 is a research project
• you need support for the latest hardware targets or production stability

Facets

library · maturity maintenance

deep-learning compiler machine-learning gpu-computing deep-learning machine-learning compilers gpu-computing cross-platform windows python cpp tensor-compiler keras-backend onnx mlir opencl portable-deep-learning linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="plaidml/plaidml")