{"full_name":"Oneflow-Inc/oneflow","owner":"Oneflow-Inc","name":"oneflow","shard":"1","url":"https://github.com/Oneflow-Inc/oneflow","description":"OneFlow is a deep learning framework designed to be user-friendly, scalable and efficient.","homepage":"http://www.oneflow.org","topics":["deep-learning","machine-learning","deep-neural-networks","ml","distributed","neural-network","cuda"],"tag_pages":["cuda","deep-learning","deep-neural-networks","distributed","machine-learning","ml","neural-network"],"language":"C++","license":"Apache-2.0","avatar":"https://avatars.githubusercontent.com/u/24632470?v=4","created_at":"2017-02-11T06:09:53Z","pushed_at":"2025-12-04T01:29:28Z","metrics":{"stars":9434,"forks":1013,"watchers":361,"issues":1067,"prs":9602,"releases":34,"issues_open":469,"issues_closed":598,"prs_open":181,"prs_closed":1732,"prs_merged":7689,"discussions":13},"growth":{"day":{"stars":0,"forks":0,"base_day":"2026-11-15"}},"history":{"dates":["2026-11-10","2026-11-11","2026-11-12","2026-11-13","2026-11-14","2026-11-15","2026-11-16"],"stars":[9436,9436,9437,9437,9435,9434,9434],"forks":[1013,1013,1013,1013,1013,1013,1013],"watchers":[null,361,361,361,361,361,361],"issues_open":[null,469,469,469,469,469,469],"prs_open":[null,180,180,180,181,181,181]},"boards":{},"same_owner":[],"issues":[{"number":9075,"title":"带有 BatchNorm2d 的模型在开启 amp 和 grad acc 时会报错","comments":29,"created_at":"2022-09-09T04:58:53Z"},{"number":3516,"title":"F0828 13:01:44.340785  7127 job_completer.cpp:71] Check failed: mutable_consumer == nullptr ","comments":14,"created_at":"2020-08-28T13:08:00Z"},{"number":7020,"title":"Consistent tensor indexing","comments":12,"created_at":"2021-12-13T09:17:18Z"},{"number":3784,"title":"数据集加载","comments":12,"created_at":"2020-11-09T04:58:10Z"},{"number":8881,"title":"为算子添加性能测试（profile）","comments":11,"created_at":"2022-08-09T02:50:33Z"},{"number":3882,"title":"Check failed: job().job_conf().train_conf().has_model_update_conf()","comments":11,"created_at":"2020-11-21T04:31:19Z"},{"number":9398,"title":"RuntimeError: This package is a placeholder. Please install oneflow following the instructions","comments":10,"created_at":"2022-11-09T02:50:01Z"},{"number":8910,"title":"NeuS网络训练速度慢","comments":10,"created_at":"2022-08-12T09:24:52Z"}],"discussions":[{"number":10319,"title":"Multi graphics card training error","comments":0,"category":"Feature Request","created_at":"2023-08-23T07:17:45Z"},{"number":10222,"title":"Anything from oneflow for new mov_to_mov","comments":0,"category":"Feature Request","created_at":"2023-05-03T06:48:09Z"},{"number":9849,"title":"OneEmbedding的推理方案是什么？","comments":1,"category":"Q&A","created_at":"2023-02-09T07:48:59Z"},{"number":9848,"title":"[ROCm] ROCm support?","comments":2,"category":"Q&A","created_at":"2023-02-08T15:31:47Z"},{"number":9450,"title":"需求：在macOs/ios/windows下可用oneflow推理加速","comments":0,"category":"Feature Request","created_at":"2022-11-20T17:06:24Z"},{"number":9326,"title":"CUTLASS Integration","comments":1,"category":"Ideas","created_at":"2022-10-27T16:42:48Z"},{"number":9272,"title":"学习了下persistent table相关实现，但发现没有compaction机制，一直顺序追加写磁盘空间不会爆嘛","comments":1,"category":"Q&A","created_at":"2022-10-19T09:03:05Z"},{"number":9058,"title":"Users Questions from Wechat","comments":1,"category":"Q&A","created_at":"2022-09-06T03:34:47Z"}],"latest_release":{"name":"Version 1.0.0 ","tag":"v1.0.0","published_at":"2024-03-11T03:18:18Z","body":"# Version 1.0.0\r\n\r\n# OneFlow v1.0.0 release note\r\n\r\nOneFlow v1.0.0 came out, welcome to install the new version for a better experience.\r\n\r\n- Highlights\r\n- New Features\r\n- Improvements\r\n- Changes and Fixes\r\n- Performance\r\n\r\n# Highlights\r\n\r\nThis version update includes 447 commits and the following highlights:\r\n\r\n- Released a new interface `compile_from_torch`. This interface, while sharing the parameter memory, converts a PyTorch Module instance into a OneFlow Module instance. It supports direct Eager execution or conversion into a static graph nn.Graph, further accelerating the process using MLIR compilation. This interface is rapidly evolving and currently supports dynamic shape compilation, validated across typical models such as ResNet50, Faster RCNN, and Stable Diffusion.\r\n\r\n- Made a series of optimizations and refactoring to Eager execution runtime, including unification of system memory pools, integration with CUDA native interfaces, optimization of instruction scheduling mechanisms, introduction of an instruction fusion mechanism, optimization of Autograd graph construction speed, optimization of Op inference process, and decoupling of Instruction and Stream, etc.\r\n\r\n- The static graph distributed physical execution plan supports separate compilation functionality, allowing each process to independently compile its required execution plan, eliminating linear growth of compilation time with GPU scale.\r\n\r\n- Addition of a series of functional automatic differentiation related interface supports, including jvp, vjp, hvp, vhp, jacobian, and hessian.\r\n\r\n- Addition of the Insight module, supporting visualization of kernel invocation, execution time, speed, and other related information within the embedded point intervals.\r\n\r\n- Updates to LiBai (the open-source toolbox for large-scale model training), with native support for fine-tuning and distributed inference of Llama2 and ChatGLM2, supporting full finetune, adapter finetune, lora finetune. lm-eval-harness can be used for language model evaluation and validation.\r\n\r\n- Upgrade of OneFlow Serving functionality, adding support for OneFlow Python backend and OneFlow Lite backend, in addition to the existing support for OneFlow Cpp backend.\r\n\r\n# New Features\r\n\r\n## 1. compile\\_from\\_torch\r\n\r\nThe `compile_from_torch` interface, while sharing the parameter memory, converts a PyTorch Module instance into a OneFlow Module instance. It supports direct Eager execution or conversion into a static graph nn.Graph, further accelerating the process using MLIR compilation. (https://github.com/Oneflow-Inc/oneflow/pull/10404, https://github.com/Oneflow-Inc/oneflow/pull/10408, https://github.com/Oneflow-Inc/oneflow/pull/9984, https://github.com/Oneflow-Inc/oneflow/pull/9754)\r\n\r\nInterface Signature and Parameter Introduction:\r\n\r\n```\r\ncompile_from_torch(torch_module: torch.nn.Module, \\*, use_graph=True, options={})\r\n* torch_module: The Torch Module instance to be converted.\r\n* use_graph: Indicates whether to transform into a static graph nn.Graph and utilize MLIR compilation acceleration. The default is True.\r\n* options:\r\n  * size: When using static graph nn.Graph, the hash value of the graph corresponding to the input shape will be calculated and cached. Size indicates the maximum capacity of the static graph cache. When exceeding the maximum capacity, the graph will be cleared based on the LRU strategy. The default value is 9.\r\n  * dynamic: For the first input with a dynamic shape, the graph will be fully compiled. For subsequent inputs with different shapes, if dynamic is True, shared graph will be used for compilation acceleration; if dynamic is False, the compilation will be performed each time. The default is True.\r\n  * debug: Debug mode and log level settings. -1 disables debug mode, 0 outputs warnings and static graph construction information, 1 additionally outputs graph construction information for each sub-module, 2 additionally outputs progress for each operator, 3 provides more detai","is_prerelease":false},"readme":true,"readme_zh":false,"news":[],"stale":false,"snapshot_at":"2026-11-16T09:23:14Z","history_start":"2026-09-06"}