sparsity

Modules

modelopt.torch.sparsity.attention_sparsity	Extensible sparse attention optimization for transformer models.
modelopt.torch.sparsity.state_sparsity	Decay-aware sparsity policies for persisted recurrent state.
modelopt.torch.sparsity.weight_sparsity	API for weight sparsification algorithms.

API for sparsification algorithms.

This module provides access to weight, attention, and recurrent-state sparsity algorithms. For backward compatibility, weight sparsity APIs are re-exported at the module level.