save_prune_retrain_model

Filter-level sparsity or 2:4 structured sparsity API. Only either of the two sparsity features can be enabled at a time.

Channel sparsity scenario: Generate a sparsity model that implements channel cropping based on the retrained model with the mask operator, and delete the mask operator.
2:4 structured sparsity scenario: Generate the final sparsity model based on the retrained model with the 2:4 structured sparsity operator, and delete the structured sparsity operator.

None

save_prune_retrain_model(pb_model, outputs, record_file, save_path)

Option	Input/Return	Description	Restriction
pb_model	Input	PB model with the mask operator for inference. The model parameters will be restored by the retrained checkpoint file.	A string
outputs	Input	Model outputs.	A list of strings, for example, [output1,output2,...].
record_file	Input	Path and name of the file that records sparsity information.	A string
save_path	Input	Model save path. Must include the prefix of the model name, for example, *./quantized_model/model**.	A string

None

Implement the sparse PB model.

amct.save_prune_retrain_model(masked_pb_path, [operation_name_1, operation_name_2], './tmp/record.txt', './pb_model/final_model')

Parent topic: Sparsity APIs