Oneflow-Inc/oneflow

Health v2 · maintenance only

Adoption not part of the score

9428 stars · 1014 forks observed · 2026-08-28

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

OneFlow is an open-source deep learning framework written in C++ with a PyTorch-like Python API, focused on scalable and efficient distributed training. It supports n-dimensional parallelism via Global Tensor and model acceleration/deployment through a Graph Compiler.

Use cases

• train deep neural networks with a PyTorch-like API
• scale model training across multiple GPUs with automatic parallelism
• run distributed deep learning training on CUDA clusters
• compile and deploy models for faster inference
• train large models with data, tensor and pipeline parallelism

When to choose

• you need efficient multi-GPU or multi-node distributed training with minimal code changes
• you want a PyTorch-like API with built-in n-dimensional parallelism
• you are training large models where distributed performance and memory efficiency matter

When to avoid

• you need the broadest ecosystem of third-party libraries and community support (PyTorch is safer)
• you develop primarily on macOS or Windows, since OneFlow targets Linux
• your team is new to deep learning and wants the most mainstream tooling and tutorials

Facets

library · maturity active

deep-learning machine-learning gpu-computing llm-training deep-learning machine-learning gpu-computing microservices python cpp pytorch-like-api distributed-training global-tensor graph-compiler cuda neural-networks model-parallelism linux gpu docker

Member repositories

For agents

markdown · JSON · MCP: product_card(name="Oneflow-Inc/oneflow")