howtospark
Training

SmoothQuant

QuantizationReference

Migrates activation outliers into the weights to unlock W8A8.

Relieves
ComputeBandwidth
Targets
Weights · Activations
Format
W8A8 (INT8)
Granularity
Per-channel scaling
Lifecycle
PTQ · no gradients
Calibration
Small calibration set
Compression
~2×
Quality
Near-FP16 at W8A8
Hardware
INT8 tensor cores
Runtimes
vLLMTensorRT-LLM

A worked Spark recipe for this method hasn't been written yet — it lives here as a reference point in the ontology.