Manual Tuning

If the accuracy after sparsity does not meet the requirements, refer to this section to perform tuning.

If the accuracy of the retrained sparsified model is not as expected, you can modify the simplified configuration file (see Simplified QAT Configuration File for details) and perform sparsity and training again. The common methods are as follows:

• Use a lower sparsity ratio. Set the prune_ratio parameter in the simplified configuration file to a smaller value and perform sparsity again.
• Skip certain layers in sparsity by setting the regular_prune_skip_layers parameter in the simplified configuration file.