coreai_opt.pruning.config.ConstantSparsitySchedule¶

class coreai_opt.pruning.config.ConstantSparsitySchedule[source]¶

Step function: zero before begin_step, target_sparsity at and after.

begin_step¶

Step at which to switch from 0 to target_sparsity. Default: 0.

compute_sparsity(step_count, target_sparsity, prev_sparsity=None)[source]¶

Return the sparsity that should be applied at step_count.

Parameters:

step_count (int) – The current step count of the pruner (monotonically increasing).
target_sparsity (float) – The final sparsity we want to reach at the end of the pruning schedule.
prev_sparsity (float | None) – Sparsity from the previous invocation. Schedules that don’t need this can ignore it; schedules that do (e.g. PolynomialDecaySchedule with an update_frequency gap) raise ValueError when omitted.

Returns:

The sparsity level to apply at the current step.

Return type:

float