Requirements

• Python 3.9 or higher
• NumPy (installed automatically)

Basic Installation

From PyPI

pip install holovec

From Source

git clone https://github.com/Twistient/HoloVec.git
cd HoloVec
pip install -e .

Optional Backends

HoloVec supports three computational backends. NumPy is always available; PyTorch and JAX are optional.

PyTorch (GPU Support)

pip install holovec[torch]
# or
pip install torch

Enables CUDA (NVIDIA) and MPS (Apple Silicon) GPU acceleration.

JAX (JIT Compilation)

pip install holovec[jax]
# or
pip install jax jaxlib

Enables JIT compilation for 10-100x speedup on repeated operations.

All Backends

pip install holovec[all]

Development Installation

For contributing or running tests:

git clone https://github.com/Twistient/HoloVec.git
cd HoloVec
pip install -e .[dev]

This includes: - pytest for testing - black for code formatting - ruff for linting - mypy for type checking - hypothesis for property-based testing

Verifying Installation

from holovec import VSA

# Check available backends
print(VSA.backend_info())

# Create a model and test basic operation
model = VSA.create('FHRR', dim=1024)
a, b = model.random(), model.random()
c = model.bind(a, b)
print(f"Binding works: {model.similarity(model.unbind(c, b), a) > 0.99}")

Expected output:

{'numpy': True, 'torch': True/False, 'jax': True/False}
Binding works: True

Backend Selection

HoloVec automatically selects NumPy as the default backend. To use a specific backend:

# NumPy (default, always available)
model = VSA.create('FHRR', dim=2048)

# PyTorch with CPU
model = VSA.create('FHRR', dim=2048, backend='torch')

# PyTorch with CUDA GPU
model = VSA.create('FHRR', dim=2048, backend='torch', device='cuda')

# PyTorch with Apple Silicon GPU
model = VSA.create('FHRR', dim=2048, backend='torch', device='mps')

# JAX with JIT compilation
model = VSA.create('FHRR', dim=2048, backend='jax')

Troubleshooting

PyTorch not detected

Ensure PyTorch is installed:

python -c "import torch; print(torch.__version__)"

JAX not detected

Ensure JAX and jaxlib are installed:

python -c "import jax; print(jax.__version__)"

CUDA not available

Check CUDA installation:

import torch
print(torch.cuda.is_available())
print(torch.cuda.get_device_name(0) if torch.cuda.is_available() else "No CUDA")

Apple Silicon MPS issues

MPS support requires macOS 12.3+ and PyTorch 1.12+:

import torch
print(torch.backends.mps.is_available())