README.md

Hybrid-Pruning Schedules

The examples in this directory show hybrid pruning schedules in which we combine several different pruning strategies.

1. alexnet.schedule_agp_2Dreg.yaml

This example presents a pruning-schedule that performs element-wise (fine grain) pruning, with 2D group (kernel) regularization. The regularization "pushes" 2D kernels towards zero, while the pruning attends to individual weights coefficients. The pruning schedule is driven by AGP.

2. alexnet.schedule_sensitivity_2D-reg.yaml

This example also presents a pruning-schedule that performs element-wise (fine grain) pruning, with 2D group (kernel) regularization. However, the pruner is a `Distiller.pruning.SensitivityPruner` which is driven by the tensors' sensitivity, instead of AGP.

Experiment	Model	Sparsity	Top1	Baseline Top1
1	Alexnet	88.31	56.40	56.55
2	Alexnet	88.31	56.24	56.55