Home

Welcome to the TorchQuery Wiki

TorchQuery is a high-performance, hardware-agnostic vectorized engine built on top of PyTorch. It is engineered to perform lightning-fast data cleaning, outlier mitigation, and neural healing directly on standard multi-core CPUs using vectorized SIMD acceleration.

🚀 Quick Start (CPU-Optimized)

Ensure you have your environment ready, then run a standard sweep:

import torch
import torchquery as tq

# Maximize multi-core CPU performance
torch.set_num_threads(4)

# Create a sample tensor with an unstable anomaly (outlier)
data = torch.randn(1000, 100)
data[50, 50] = 999.0  # Anomaly

# Run the streaming SDC mitigation engine
healed_data = tq.SDCEngine.protect(data, sigma=3.0)
print("Tensor successfully healed and stabilized!")