faissknn

faissknn contains implementations for both multiclass and multilabel K-Nearest Neighbors Classifier implementations. The classifiers follow the scikit-learn: fit, predict, and predict_proba methods.

Install

faissknn needs a FAISS backend, which you choose at install time via an extra. Pick exactly one — they ship the same faiss module and can't coexist. For CPU (works everywhere — macOS, Windows, Linux; no GPU, CUDA driver, or system toolkit required):

pip install "faissknn[cpu]"

This pulls in faiss-cpu along with numpy and torch.

A bare pip install faissknn installs no FAISS backend and raises a clear error at import time telling you to pick an extra. Always install one of [cpu] or [cuda].

GPU acceleration

On Linux x86_64 with an NVIDIA driver (R525+), use the [cuda] extra, which installs faiss-cuda (Taylor Geospatial's GPU wheels, CUDA 12.8, arch sm_70/80/86/89/90: V100, A100, A10/A30/RTX-30, RTX-40, H100/H200 — T4/Blackwell/RTX-50 pending a PyPI size-limit increase) instead of faiss-cpu. No system CUDA toolkit needed — the runtime libraries come from nvidia-cuda-runtime-cu12 / nvidia-cublas-cu12 on PyPI. The GPU wheel contains the full CPU implementation too, so it also works on GPU-less machines.

pip install "faissknn[cuda]"

uv/pip can't auto-detect the host CUDA driver, so the backend is a manual choice. Because nothing is installed until you pick an extra, a fresh install of any single extra is clean — no base faiss-cpu to fight, no uninstall/reinstall dance. If you later want to switch backends in the same environment, uninstall the current one first (e.g. pip uninstall -y faiss-cpu) before installing the other extra, since the FAISS packages share the faiss module and can't coexist.

Usage

Multiclass:

from sklearn.datasets import make_classification
from sklearn.model_selection import train_test_split

from faissknn import FaissKNNClassifier

x, y = make_classification()
x_train, x_test, y_train, y_test = train_test_split(x, y)
model = FaissKNNClassifier(
    n_neighbors=5,
    n_classes=None,
    device="cpu"
)
model.fit(x_train, y_train)

y_pred = model.predict(x_test) # (N,)
y_proba = model.predict_proba(x_test) # (N, C)

Multilabel:

from sklearn.datasets import make_classification
from sklearn.model_selection import train_test_split

from faissknn import FaissKNNMultilabelClassifier

x, y = make_multilabel_classification()
x_train, x_test, y_train, y_test = train_test_split(x, y)
model = FaissKNNMultilabelClassifier(
    n_neighbors=5,
    device="cpu"
)
model.fit(x_train, y_train)

y_pred = model.predict(x_test) # (N, C)
y_proba = model.predict_proba(x_test) # (N, C)

GPU/CUDA: faissknn also supports running on the GPU to speed up computation. Simply change the device to cuda or a specific cuda device cuda:0

model = FaissKNNClassifier(
    n_neighbors=5,
    device="cuda"
)
model = FaissKNNClassifier(
    n_neighbors=5,
    device="cuda:0"
)

Cite

If you use faissknn in your research, please considering citing!

@software{isaac_corley_2026_18370748,
  author       = {Isaac Corley},
  title        = {isaaccorley/faissknn: Zenodo Cite},
  month        = jan,
  year         = 2026,
  publisher    = {Zenodo},
  version      = {v0.0.3},
  doi          = {10.5281/zenodo.18370748},
  url          = {https://doi.org/10.5281/zenodo.18370748},
}