Quantize

Quantize the sub-modules of a module according to a predicate.

Abstract Signature:

Quantize(model: Module, group_size: int = 64, bits: int = 4)

PyTorch

API: torch.ao.quantization.quantize_dynamic
Strategy: Direct Mapping
Official Docs ↗

JAX (Core)

API: —
Strategy: Plugin (quantization)

Keras

API: keras.layers.ZeroPadding3D.quantize
Strategy: Plugin (quantization)

TensorFlow

API: —
Strategy: Plugin (quantization)

Apple MLX

API: mlx.nn.quantize
Strategy: Direct Mapping
Official Docs ↗

Flax NNX

API: —
Strategy: Plugin (quantization)

PaxML / Praxis

API: —
Strategy: Plugin (quantization)