diff --git a/README.md b/README.md index 4a3fcee..9ba5462 100644 --- a/README.md +++ b/README.md @@ -1,37 +1,190 @@ +# DuaLip: Dual Decomposition-based Linear Program Solver -# PyTorch Rewrite - Temporary README +[![License](https://img.shields.io/badge/License-BSD_2--Clause-blue.svg)](LICENSE) +[![Python 3.10+](https://img.shields.io/badge/python-3.10%2B-blue.svg)](https://www.python.org/downloads/) +[![PyTorch 2.6+](https://img.shields.io/badge/PyTorch-2.6%2B-ee4c2c.svg)](https://pytorch.org/) +[![Docs](https://img.shields.io/badge/docs-GitHub_Pages-green.svg)](https://linkedin.github.io/DuaLip) -The branch `dualip-pytorch` serves as the master branch while we update the repo -to the new Pytorch version of DuaLip. This branch is protected, and so changes -should be made by creating a new branch from `dualip-pytorch` and then submitting -a pull request. +DuaLip is an **extreme-scale Linear Program (LP) solver** built on PyTorch. It solves structured LP problems of the form: -Once `dualip-pytorch` is stable enough for an initial release, we will set -`dualip-pytorch` to be the master branch. The older scala based master branch -is frozen under the `dualip-legacy-scala` branch. +``` +minimize c^T x +subject to Ax ≤ b + x_i ∈ C_i for all i ∈ {1, 2, ..., I} +``` + +where `x = (x_1, ..., x_I)` is the full vector of optimization variables, `x_i` is the vector associated with entity `i` (e.g., a user), and `A, b, c, C_i` are problem-specific data. + +Many large-scale decision systems — allocation, assignment, marketplace shaping — reduce to solving very large LPs on a recurring cadence. General-purpose LP solvers treat the constraint matrix as an unstructured operator and have not scaled to these extreme problem sizes. Specialized matching solvers cannot accommodate the heterogeneous constraint families (budgets, pacing, fairness, frequency caps) that arise in production. + +DuaLip fills this gap with a **ridge-regularized dual ascent** approach that: +- Exploits block-diagonal structure for massive parallelism +- Requires no commercial solver license +- Runs entirely on commodity GPUs with open-source PyTorch + +## Features + +- **Extreme scale** — structured LPs with hundreds of millions of entities and up to trillions of variables +- **GPU-accelerated** — significant wall-clock speedups over the prior Spark-based solver through native CUDA execution +- **Operator-centric API** — new LP formulations require only a new `ObjectiveFunction`; projections and the solve loop are reused +- **Distributed multi-GPU** — scales horizontally via `torch.distributed` (NCCL); communication cost depends only on the dual dimension +- **Nesterov-accelerated gradient ascent** — with Lipschitz-based step sizing, Jacobi preconditioning, and optional γ decay +- **Polytope projections** — box, simplex, and cone constraints, extensible via a registry pattern +- **Warm start** — initialize from a previous dual solution for faster convergence on recurring workloads +- **MLflow integration** — track dual objective, step sizes, hyperparameters, and convergence diagnostics +- **Python-native** — aligns with common numerical and ML stacks for easy instrumentation and debugging +## Quick start +### Requirements -# Development setup and workflow +- Python >= 3.10 +- PyTorch >= 2.6.0 -Clone the Github repo and then create a new branch from `dualip-pytorch`. +### Installation -1) Create & activate a virtual environment ```bash -python -m venv .venv && source .venv/bin/activate # Windows: .\.venv\Scripts\activate +pip install -e . ``` -2) Install dev dependencies and pre-commit hooks -```bash -make install +### Usage + +```python +import torch +from dualip.objectives.matching import MatchingInputArgs +from dualip.projections.base import create_projection_map +from dualip.run_solver import run_solver +from dualip.types import ComputeArgs, ObjectiveArgs, SolverArgs + +# Build your problem data (A and c in CSC format, budget vector b) +# and a projection map defining per-column constraints +input_args = MatchingInputArgs( + A=A_csc, # sparse constraint matrix (CSC) + c=C_csc, # sparse cost matrix (CSC) + b_vec=b_vec, # budget / RHS vector + projection_map=create_projection_map(projection_entries), +) + +solver_args = SolverArgs(max_iter=5000, gamma=1e-3) +compute_args = ComputeArgs(host_device="cuda:0") +objective_args = ObjectiveArgs(objective_type="matching") + +result = run_solver(input_args, solver_args, compute_args, objective_args) + +print(f"Dual objective: {result.dual_objective:.6f}") ``` -3) Run tests -```bash -make test +For a complete end-to-end example, see [examples/movielens_matching/movies_lens_matching.py](examples/movielens_matching/movies_lens_matching.py) which solves a matching problem on the MovieLens dataset. An example using a MIPLIB 2017 problem is available under [examples/miplib_2017/](examples/miplib_2017/). + +For reference on setting up your own problem, see [benchmark/run_matching_benchmark.py](benchmark/run_matching_benchmark.py) (single-GPU) and [benchmark/run_matching_benchmark_dist.py](benchmark/run_matching_benchmark_dist.py) (distributed multi-GPU). + +## PyTorch-based solver + +DuaLip was originally built on Scala/Spark. The PyTorch rewrite is a ground-up redesign that replaces the CPU-centric, schema-bound runtime with a GPU-native, composable architecture, delivering significant wall-clock speedups while making new problem formulations far easier to express. + +The solver uses an **operator-centric programming model** built around three decoupled primitives — `ObjectiveFunction`, `ProjectionMap`, and `Maximizer` — so new LP formulations only require implementing a new `ObjectiveFunction` while reusing the solve loop, projections, and distributed primitives. + +**Key design choices for GPU performance:** +- Sparse CSC storage aligned with block-diagonal structure +- Batched projection kernels +- Jacobi preconditioning with γ continuation schedule +- Distributed multi-GPU via `torch.distributed` (NCCL) +- Warm start support +- Pure PyTorch — no custom C++/CUDA kernels + +## Documentation + +Full documentation is available at **[linkedin.github.io/DuaLip](https://linkedin.github.io/DuaLip)**. + +## Migrating from the Spark-based solver + +This PyTorch-based solver is a ground-up rewrite of the original [Spark/Scala DuaLip solver](https://arxiv.org/abs/2103.05277). The core dual-decomposition algorithm is the same, but the implementation, data formats, and APIs are entirely new. + +**Version history:** Tags `v1.0.0` through `v4.0.5` correspond to the Spark/Scala solver. The PyTorch rewrite begins at `v5.0.1` on the `master` branch. + +See the [Quick start](#quick-start) section and [examples/movielens_matching/movies_lens_matching.py](examples/movielens_matching/movies_lens_matching.py) for examples of the new API. + +## Development + +1. **Create and activate a virtual environment** + + ```bash + python -m venv .venv && source .venv/bin/activate + ``` + +2. **Install dev dependencies and pre-commit hooks** + + ```bash + make install + ``` + +3. **Run tests** + + ```bash + make test + ``` + +4. **Run checkstyle** (format + lint) + + ```bash + make checkstyle + ``` + +## Repository structure + +``` +src/dualip/ Core library + objectives/ Objective functions (matching, MIPLIB) + optimizers/ Accelerated gradient descent solver + projections/ Box, simplex, cone projections + preprocessing/ Input validation and Jacobi preconditioning + utils/ Sparse, distributed, and MLflow utilities +tests/ Unit and distributed tests +examples/ Example problems (MovieLens matching, MIPLIB 2017) +benchmark/ Scaling and performance benchmarks +docs/ Sphinx documentation source ``` -4) Run checkstyle (format + lint) -```bash -make checkstyle -``` \ No newline at end of file +## Citing DuaLip + +If you use DuaLip in your work, please cite: + +```bibtex +@inproceedings{ramanath:21, + author = {Ramanath, Rohan and Keerthi, Sathiya S. and Basu, Kinjal and Salomatin, Konstantin and Yao, Pan}, + title = {Efficient Algorithms for Global Inference in Internet Marketplaces}, + journal = {arXiv preprint arXiv:2103.05277}, + year = {2021}, + url = {https://arxiv.org/abs/2103.05277} +} + +@InProceedings{pmlr-v119-basu20a, + title = {{ECLIPSE}: An Extreme-Scale Linear Program Solver for Web-Applications}, + author = {Basu, Kinjal and Ghoting, Amol and Mazumder, Rahul and Pan, Yao}, + booktitle = {Proceedings of the 37th International Conference on Machine Learning}, + pages = {704--714}, + year = {2020}, + volume = {119}, + series = {Proceedings of Machine Learning Research}, + publisher = {PMLR}, + url = {http://proceedings.mlr.press/v119/basu20a.html} +} +``` + +## Contributing + +Please report bugs and feature requests via the [GitHub issue tracker](https://github.com/linkedin/DuaLip/issues). To contribute code, see the [contributing guide](https://linkedin.github.io/DuaLip/contributing/index.html). + +## Acknowledgements + +DuaLip is built and maintained by the AI Algorithms team at LinkedIn: + +- [Gregory Dexter](https://www.linkedin.com/in/gregorydexter1/) +- [Aida Rahmattalabi](https://www.linkedin.com/in/aida-rahmattalabi-23a4ab4a/) +- [Sanjana Garg](https://www.linkedin.com/in/sanjana-garg/) +- [Qingquan Song](https://www.linkedin.com/in/qingquan-song-b71167119/) +- [Zhipeng Wang](https://www.linkedin.com/in/zhipeng-wang-phd-66806816/) + +## License + +BSD 2-Clause License. See [LICENSE](LICENSE) for details. diff --git a/docs/demo/matching.rst b/docs/demo/matching.rst index 7e06112..e6fe524 100644 --- a/docs/demo/matching.rst +++ b/docs/demo/matching.rst @@ -117,7 +117,7 @@ The solver can be run locally with the following command: .. code:: bash - python demo/movies_lens_matching.py --out_prefix data/movielens/ --run_solver --ratings_csv_path data/ml-20m/data/ratings.csv + python examples/movielens_matching/movies_lens_matching.py --out_prefix data/movielens/ --run_solver --ratings_csv_path data/ml-20m/data/ratings.csv This will save the input tensors and projection maps to the directory specified by :code:`--out_prefix`. It will also run the solver and save the results to the directory specified by :code:`--out_prefix`. diff --git a/demo/logs/movies_log.txt b/examples/movielens_matching/logs/movies_log.txt similarity index 100% rename from demo/logs/movies_log.txt rename to examples/movielens_matching/logs/movies_log.txt diff --git a/demo/logs/movies_with_fairness_log.txt b/examples/movielens_matching/logs/movies_with_fairness_log.txt similarity index 100% rename from demo/logs/movies_with_fairness_log.txt rename to examples/movielens_matching/logs/movies_with_fairness_log.txt diff --git a/demo/movies_lens_matching.py b/examples/movielens_matching/movies_lens_matching.py similarity index 100% rename from demo/movies_lens_matching.py rename to examples/movielens_matching/movies_lens_matching.py