Workflow

This section describes the sparsified layers supported by manual sparsity, and API call sequence and example.

Currently, AMCT supports only retraining-based filter-level sparsity. For the sparsity example, see "Filter-level sparsity" in Sample List. The layers supported by filter-level sparsity as well as their constraints are listed as follows.

**Table 1** Layers supported by filter-level sparsity and their constraints
Technique	Supported Layer Type	Constraints
Filter-level sparsity	torch.nn.Linear	Layers sharing the weight and bias parameters do not support sparsity.
Filter-level sparsity	torch.nn.Conv2d	Layers sharing the weight and bias parameters do not support sparsity. Depthwise processing supports only passive sparsity (groups = in_channels). The shape of the input data must be (N, Cin, Hin, Win).

API Call Sequence

The following figure shows the API call sequence of filter-level sparsity. The training runs on the CPU or NPU environment of the PyTorch framework. Based on the inference script of the open-source framework, the AMCT API is called to compress the model. The compressed model needs to be converted into an offline model that adapts to the AI processor using the ATC before it can be used for inference on the AI processor.

Figure 1 API call sequence for filter-level sparsity

The user implements the operations in blue, while those in gray are implemented by using AMCT APIs.

Build an original PyTorch model, call the create_prune_retrain_model API to modify the model, and insert the filter-level sparsity mask operator into the graph structure to prune model parameters.
Train the modified model until the accuracy meets your requirement. If the training is interrupted, call the restore_prune_retrain_model API to prune the original model again based on the sparsity record file and perform QAT until the accuracy meets your requirement.
Generate the .pth file that meets the accuracy requirement based on the final retrained filter-level sparsity model. Alternatively, call the save_prune_retrain_model API to generate the final ONNX fake-quantized model and deployable model.

Example

Take the following steps to get started. Update the sample code based on your situation.
Tweak the arguments passed to AMCT API calls as required. Sparsity relies on the user training result. Ensure that a PyTorch training script that yields satisfactory training accuracy is available.

Import the AMCT package and set the log level using the environment variable in "AMCT (PyTorch)" in Post-installation Actions.
1
import amct_pytorch as amct
(Optional) Run inference on the original model in the PyTorch environment based on the test dataset to validate the environment setup and inference script. (Update the sample code based on your situation.)
This step is recommended as it guarantees a properly functioning original model for inference with acceptable accuracy. You can use a subset from the test dataset to improve the efficiency.
1 2 3
ori_model.load() # Test the model. user_test_model(ori_model, test_data, test_iterations)

Run AMCT to perform training with the sparsity operator.

Modify the original model and insert the filter-level sparsity mask operator into the graph structure.

Before performing this step, restore the already trained parameters, for example, ori_model.load() in 2.

simple_cfg = './retrain.cfg'
record_file = './tmp/record.txt'
prune_retrain_model = amct.create_prune_retrain_model(model=ori_model,
                                input_data=ori_model_input_data,
                                config_defination=simple_cfg,
                                record_file=record_file)

Implement gradient descent optimization on the modified graph and train the graph on the training dataset. (Update the sample code based on your situation.)
1. Implement gradient descent optimization on the modified graph.
  Perform this step after the model is trimmed.
  1
  optimizer = user_create_optimizer(prune_retrain_model)
2. Restore the model from existing checkpoints and train the model.
  Note: Restore model parameters from checkpoints before training.
  1 2
  quant_pth = './ckpt/user_model' user_train_model(optimizer, prune_retrain_model, train_data)
(Optional) If the save_prune_retrain_model API is called, refer to this step. If the model is saved as a .pth file, skip this step.
Save the model to implement filter-level sparsity.
1 2 3 4
prune_retrain_model = amct.save_prune_retrain_model( model=pruned_retrain_model, save_path=save_path, input_data=input_data)

(Optional) Run inference on the sparsified model (prune_retrain_model) in the ONNX Runtime environment based on the test dataset (test_data) to test the accuracy. (Update the sample code based on your situation.)
Check the accuracy drop (from filter-level sparsity) of the fake-quantized model after sparsity by comparing with that of the original model in 2.
1 2
prune_retrain_model = './results/user_model_fake_prune_model.onnx' user_do_inference_onnx(prune_retrain_model, test_data, test_iterations)

If the training is interrupted, restore data from the checkpoints to resume the training.

Import the AMCT package and set the log level using the environment variable in "AMCT (PyTorch)" in Post-installation Actions.
1
import amct_pytorch as amct

Prepare an original model.
1
ori_model= user_create_model()

Run AMCT to resume the QAT process.

Modify the model, insert the filter-level sparsity mask operator into the graph structure, and save the model as a new prune_model.

model = ori_model
input_data = ori_model_input_data
record_file = './tmp/record.txt'
config_defination = './prune_cfg.cfg'
save_pth_path = /your/path/to/save/tmp.pth
model.load_state_dict(torch.load(state_dict_path))
prune_retrain_model = amct.restore_prune_retrain_model(model=ori_model,
                                                       input_data=ori_model_input_data,
                                                       record_file=record_file,
                                                       config_defination='./prune_cfg.cfg',
                                                       save_pth_path=/your/path/to/save/tmp.pth,
                                                       'state_dict')

Implement gradient descent optimization on the modified model, restore parameters from the checkpoints, and train the model on the training dataset. (Update the sample code based on your situation.)
1. Restore the model parameters from the checkpoints after sparsity.
  1 2
  quant_pth = './ckpt/user_prune_model' user_train_model(optimizer, prune_retrain_model, train_data)
2. Implement gradient descent optimization on the modified graph.
  Perform this step after model parameters are restored.
  1
  optimizer = user_create_optimizer(prune_retrain_model)
3. Restore the model from existing checkpoints and train the model.
  Note: Restore model parameters from checkpoints before training.
  1
  user_train_model(optimizer, prune_retrain_model, train_data)
(Optional) If the save_prune_retrain_model API is called, refer to this step. If the model is saved as a .pth file, skip this step.
Save the model to implement filter-level sparsity.
1 2 3 4
prune_retrain_model = amct.save_prune_retrain_model( model=pruned_retrain_model, save_path=save_path, input_data=input_data)

(Optional) Run inference on the sparsified model (prune_retrain_model) in the ONNX Runtime environment based on the test dataset (test_data) to test the accuracy. (Update the sample code based on your situation.)
Check the accuracy drop (from filter-level sparsity) of the fake-quantized model after sparsity by comparing with that of the original model in 2.
1 2
prune_retrain_model = './results/user_model_fake_prune_model.onnx' user_do_inference_onnx(prune_retrain_model, test_data, test_iterations)

Follow-up Operations

If the sparsified model is output in .pth format, refer to this section. If the save_prune_retrain_model API is called, skip this section.

The output .pth model is not directly deployable for inference. Before using ATC to convert the model, you need to convert the .pth model into an ONNX network model, or call the save_prune_retrain_model API to save the .pth model as the final ONNX fake-quantized model and deployable model. The following is an example of calling the save_prune_retrain_model API:

prune_retrain_model = amct.10.6.3-save_prune_retrain_model(
     model=pruned_retrain_model,
     save_path=save_path,
     input_data=input_data)

Parent topic: Manual Sparsity