guardex/toxicity-fast-onnx

ONNX-quantized toxicity classifier used by the GuardEx SDK as the Layer-2 fast gate in its safety-classification cascade.

Files

File Use
model_int8.onnx INT8 quantized โ€” default, ~3-5x faster
model.onnx FP32 โ€” full precision, use when accuracy matters more than latency

Labels

{0: "neutral", 1: "toxic"}

Usage via GuardEx SDK

from guardex import Guard
guard = Guard()  # pulls this model on first call, caches under ~/.cache/huggingface
Downloads last month
3
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support