Installation

Moreau is open source under the Apache 2.0 license and is distributed via PyPI. Install with pip:

pip install moreau

---

Quick Install

CPU Only

pip install moreau

Pure Python interface with Rust-based CPU solver.

GPU Support

pip install moreau[cuda]

CUDA 12 by default, or moreau[cuda13] for CUDA 13. Requires Python 3.12+.

---

Choose Your Framework

PyTorch

If PyTorch (>= 2.4) is installed, moreau.torch works automatically — no extra install step needed.

Provides moreau.torch.Solver for neural network integration with full autograd support.

JAX

If JAX (>= 0.4.0) is installed, moreau.jax works automatically — no extra install step needed.

Provides moreau.jax.Solver compatible with jax.grad, jax.vmap, and jax.jit.

---

Available Extras

Extra	Installs	Notes
moreau[cuda]	moreau-cuda12	CUDA 12 GPU backend (default)
moreau[cuda12]	moreau-cuda12	CUDA 12 GPU backend (explicit)
moreau[cuda13]	moreau-cuda13	CUDA 13 GPU backend
moreau[test]	pytest, cvxpy	Test dependencies

---

Supported Platforms

OS	Arch	CPU	CUDA 12	CUDA 13
Linux	x86_64	✓	✓	✓
Linux	aarch64	✓	✓	✓
macOS	ARM64	✓	—	—

Linux wheels target manylinux_2_28 (glibc ≥ 2.28). aarch64 CUDA wheels cover Grace Hopper (sm_90) and Orin/Thor (sm_87); CUDA 13 aarch64 additionally covers Blackwell (sm_120). If your device’s compute capability is not listed, build from source with MOREAU_CUDA_ARCH=<capability>.

---

Verify Installation

Run the built-in diagnostic to check that everything is working:

python -m moreau check

This tests all installed backends (CPU, CUDA), framework integrations (PyTorch, JAX), and autograd support in a single command.

For manual verification with each framework:

Basic

import moreau
import numpy as np
from scipy import sparse

# Simple QP
P = sparse.diags([1.0, 1.0], format='csr')
q = np.array([1.0, 1.0])
A = sparse.csr_array([[1.0, 0.0], [0.0, 1.0]])
b = np.array([0.5, 0.5])

cones = moreau.Cones(num_nonneg_cones=2)
solver = moreau.Solver(P, q, A, b, cones=cones)
solution = solver.solve()

print(f"Solution: {solution.x}")
print(f"Status: {solver.info.status}")
print("Installation successful!")

PyTorch

import torch
from moreau.torch import Solver
import moreau

cones = moreau.Cones(num_nonneg_cones=2)
solver = Solver(
    n=2, m=2,
    P_row_offsets=torch.tensor([0, 1, 2]),
    P_col_indices=torch.tensor([0, 1]),
    A_row_offsets=torch.tensor([0, 1, 2]),
    A_col_indices=torch.tensor([0, 1]),
    cones=cones,
)

P_values = torch.tensor([1.0, 1.0], dtype=torch.float64)
A_values = torch.tensor([1.0, 1.0], dtype=torch.float64)

q = torch.tensor([1.0, 1.0], dtype=torch.float64, requires_grad=True)
b = torch.tensor([0.5, 0.5], dtype=torch.float64)
solution = solver.solve(P_values, A_values, q, b)

print(f"Solution: {solution.x}")
print(f"Device: {solver.device}")
print("PyTorch installation successful!")

JAX

import jax.numpy as jnp
from moreau.jax import Solver
import moreau

cones = moreau.Cones(num_nonneg_cones=2)
solver = Solver(
    n=2, m=2,
    P_row_offsets=jnp.array([0, 1, 2]),
    P_col_indices=jnp.array([0, 1]),
    A_row_offsets=jnp.array([0, 1, 2]),
    A_col_indices=jnp.array([0, 1]),
    cones=cones,
)

P_data = jnp.array([1.0, 1.0])
A_data = jnp.array([1.0, 1.0])
q = jnp.array([1.0, 1.0])
b = jnp.array([0.5, 0.5])

solution = solver.solve(P_data, A_data, q, b)

print(f"Solution: {solution.x}")
print(f"Device: {solver.device}")
print("JAX installation successful!")

---

GPU Acceleration

CUDA Support

For NVIDIA GPUs, the CUDA backend provides significant speedups, especially for batched problems.

Check GPU availability:

import moreau

print(f"Available devices: {moreau.available_devices()}")
print(f"Default device: {moreau.default_device()}")
print(f"CUDA available: {moreau.device_available('cuda')}")

Force GPU usage:

settings = moreau.Settings(device='cuda')
solver = moreau.Solver(P, q, A, b, cones=cones, settings=settings)

---

Dependencies

Core Dependencies

Package	Version	Purpose
Python	>= 3.9	Runtime
NumPy	>= 1.19.0	Array operations
Pydantic	>= 2.0.0	Data validation
moreau-cpu	>= 0.3.3	CPU solver backend

Optional Dependencies

Package	Version	Purpose
moreau-cuda12	>= 0.3.3	GPU solver backend (CUDA 12, Python >= 3.12)
moreau-cuda13	>= 0.3.3	GPU solver backend (CUDA 13, Python >= 3.12)
PyTorch	>= 2.4	PyTorch integration
JAX	>= 0.4.0	JAX integration
SciPy	>= 1.6.0	Sparse matrix construction (used in examples)

---

Troubleshooting

ImportError: moreau-cpu not found

The core CPU backend is required. Install it with:

pip install moreau-cpu

Or reinstall moreau:

pip install --force-reinstall moreau

CUDA not detected

Ensure you have:

1. An NVIDIA GPU with CUDA support

2. CUDA drivers installed

3. The CUDA backend: pip install moreau[cuda]

Check CUDA availability:

import moreau
print(moreau.device_available('cuda'))

# Diagnose the error
error = moreau.device_error('cuda')
if error:
    print(f"CUDA error: {error}")

PyTorch/JAX import errors

Install the frameworks separately:

• PyTorch: See pytorch.org/get-started

• JAX: See docs.jax.dev/installation

Once installed, moreau.torch and moreau.jax are available automatically.