Created by: pengwangucla

Hi I converted a model from a tensorflow to paddle, the model size is around 120M

Saving with paddle

with gzip.open(model_name, 'w') as f:
                parameters.to_tar(f)

and load with

with gzip.open(model_name, 'r') as f:
            parameters = paddle.parameters.Parameters.from_tar(f)

For the same model, tensorflow restore just cost around 3s, and saving and loading with this method, it cost 60s. I wonder why is this, @helinwang will follow up with a test case.