Installation¶

Diffid is available as a Python package with pre-built wheels for most platforms.

Installation Methods¶

uv pip install diffid

pip install diffid

Optional Dependencies¶

Diffid has optional plotting support via matplotlib:

uv pip install "diffid[plotting]"

pip install "diffid[plotting]"

Verifying Installation¶

After installation, verify that Diffid is working correctly:

import diffid
import numpy as np

# Simple test
def test_func(x):
    return np.asarray([(x[0] - 1.0) ** 2])

builder = diffid.ScalarBuilder().with_objective(test_func).with_parameter("x", 0.0)
problem = builder.build()
result = problem.optimise()

print(f"Success: {result.success}")
print(f"Optimal x: {result.x[0]:.3f}")

Expected output:

Success: True
Optimal x: 1.000

Platform-Specific Notes¶

Linux¶

Pre-built wheels are available for x86_64 and aarch64 architectures. No additional setup required.

macOS¶

Pre-built wheels are available for both Intel (x86_64) and Apple Silicon (arm64) Macs.

If you're using Apple Silicon and encounter issues, ensure you're using a native arm64 Python installation rather than running under Rosetta.

Windows¶

Windows builds are marked experimental. Pre-built wheels are available but don't currently support diffsol gradients due to LLVM integration issues.

If you encounter issues consider using Windows Subsystem for Linux (WSL)

Building from Source¶

If you need to build from source (for development or if pre-built wheels aren't available):

Prerequisites¶

• Rust >= 1.70
• Python >= 3.11
• uv (recommended)

Build Steps¶

# Clone the repository
git clone https://github.com/bradyplanden/diffid.git
cd diffid

# Create Python environment
uv sync

# Build Rust extension with Python bindings
uv run maturin develop

# Run tests
uv run pytest -v

For additional troubleshooting, see the Troubleshooting Guide.

Next Steps¶

Now that Diffid is installed, proceed to the 5-Minute Quickstart to run your first optimisation.