TE-CCL is a tool to generate collective communication schedules for large topologies using a Traffic Engineering-based solver.
TE-CCL takes in a topology and collective (e.g. AllGather) and outputs a schedule (in JSON) detailing data transfer steps for each node that satisfies the demands specified by the collective. In a Traffic Engineering-based approach, TE-CCL encodes the collective communication process into capacity constraints, flow conservation constraints, and destination constraints, and solves a mixed-integer linear program (MILP), with options to convert to an linear program (LP) form or A* form for better scalibility.
Rethinking Machine Learning Collective Communication as a Multi-Commodity Flow Problem
Xuting Liu, Behnaz Arzani, Siva Kesava Reddy Kakarla, Liangyu Zhao, Vincent Liu, Miguel Castro, Srikanth Kandula, and Luke Marshall
SIGCOMM 2024 [https://doi.org/10.1145/3651890.3672249]
[TO BE FILLED]
- Install Anaconda and activate an anaconda environment.
- Obtain and install a Gurobi license. After getting the license, following the steps to install and activate Gurobi.
conda config --add channels http://conda.anaconda.org/gurobi
conda install -c conda-forge gurobi -y
<command to install Gurobi license. e.g. grbgetkey xxxxxx>
In the anaconda environment with Gurobi installed, run
pip install .
TE-CCL takes in a JSON user input file that specifies the topology, Gurobi settings, and model instance parameters. Please refer to Input Data for details.
teccl solve --input_args <input.json>
To generate a schedule for AllGather in the NDv2 topology, run
teccl solve --input_args teccl/examples/sample_inputs/ndv2_input.json
This will generate the schedule file teccl/examples/schedules/ndv2_schedule.json
For detailed examples, please refer to instructions in the examples directory.
A user input JSON file is consists of three parts: TopologyParams, GurobiParams, and InstanceParams. Detailed explainations of each argument is in input_data.py.
This sepcifies the topology considered. Each topology is defined as a seperate Python file in teccl/topologies.
Parameters for the Gurobi solver.
Parameters for the model instance, including the collective, the choice of objective function and size of epochs.
This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.
When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.
This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact [email protected] with any additional questions or comments.
This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.