save_prune_retrain_model

Applicability

Product

Supported

Atlas A3 training series products/Atlas A3 inference series products

  • Filter-level sparsity: √
  • 2:4 structured sparsity API: √

Atlas A2 training products/Atlas A2 inference products

  • Filter-level sparsity: √
  • 2:4 structured sparsity API: √

Atlas 200I/500 A2 inference product

  • Filter-level sparsity: √
  • 2:4 structured sparsity API: √

Atlas inference series products

  • Filter-level sparsity: √
  • 2:4 structured sparsity API: x

Atlas training products

  • Filter-level sparsity: √
  • 2:4 structured sparsity API: x

Note: For the Products marked with x, no error is reported when the API is called for the 2:4 structured sparsity feature, but the performance benefits cannot be obtained.

Description

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

  • Filter-level sparsity: Generates a sparsity model that implements channel cropping based on the retrained model with the mask operator, and removes the mask operator.
  • 2:4 structured sparsity: Generates the final sparsity model based on the retrained model with the 2:4 structured sparsity operator, and removes the structured sparsity operator.

Prototype

1
save_prune_retrain_model(pb_model, outputs, record_file, save_path)

Parameters

Parameter

Input/Output

Description

pb_model

Input

.pb model for inference with sparse operators. The parameters are restored from the retraining checkpoint.

A string.

outputs

Input

Output of the user model

A list of strings, for example, [output1,output2,...].

record_file

Input

Path (including the file 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, ./prune_model/*model.

A string.

Returns

None

Example

1
amct.save_prune_retrain_model(masked_pb_path, [operation_name_1, operation_name_2], './tmp/record.txt', './pb_model/final_model')

Flush file: .pb model for sparsity.