目录
- wikitextsource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- CODE_OF_CONDUCT.mdsource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- CONTRIBUTING.mdsource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- Cargo.tomlsource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- LICENSEInitial commit2年前
- NOTICEsource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- README.mdsource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- analyze.pysource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- collectives.pysource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- config.pysource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- ddp.pysource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- environment.ymlsource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- hap.rssource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- master.pysource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- models.pysource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- plot_collective_by_group_call.pysource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- profiler.pysource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- runsource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- run_allsource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
- worker.pysource code for HAP paper "HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis".2年前
邀请码
版权所有：中国计算机学会技术支持：开源发展技术委员会
京ICP备13000930号-9
京公网安备 11010802032778号
HAP: SPMD DNN Training on Heterogeneous GPU Clusters with Automated Program Synthesis
Build From Source
HAP is partially implemented in Rust and requires a Rust compiler to build. The building toolchain can be installed by following https://rustup.rs. Currently HAP requires a nightly build of the Rust compiler.
Run the following command to build HAP:
Installing Python Dependencies
HAP is implemented on top of PyTorch 1.13.1. It can be installed, along with its own dependencies, by following https://pytorch.org/get-started/previous-versions/#v1131.
Alternatively, we provides an environment file
environment.ymlfor reproducing the experiment environment with conda (https://conda.io).Contributing
See CONTRIBUTING.md for details.
Contributors
HAP is developed by Alibaba Group and HKU Netexplo. This work is supported by Alibaba Innovative Research(AIR).
License
HAP is licensed under the Apache License (Version 2.0). See LICENSE file. This product contains some third-party testcases under other open source licenses. See the NOTICE file for more information.