Automatic Compression

automatic_compression(self, input_model_path: str, output_dir: str, input_shapes: List[Dict[str, int]], framework: Framework = Framework.PYTORCH, compression_ratio: float = 0.5) → CompressorMetadata

Compress a model automatically based on the given compression ratio.

Parameters:

input_model_path (str) – The file path where the model is located.
output_dir (str) – The local path to save the compressed model.
input_shapes (List[Dict[str, int]]) – Input shapes of the model.
framework (Framework, optional) – The framework of the model.
compression_ratio (float, optional) – The compression ratio for automatic compression. Defaults to 0.5.

Raises:

e – If an error occurs while performing automatic compression.

Returns:

Compress metadata.

Return type:

CompressorMetadata

Details of Parameters

Compression Ratio

Note

As the compression ratio increases, you can get more lighter and faster compressed models, but with greater lost of accuracy.
Therefore, it is necessary to find an appropriate ratio for your requirements. It might require a few trials and errors.
The range of available values is as follows.

$0< ratio < 1$

Example

from netspresso import NetsPresso


netspresso = NetsPresso(email="YOUR_EMAIL", password="YOUR_PASSWORD")

compressor = netspresso.compressor_v2()
compressed_model = compressor.automatic_compression(
    input_shapes=[{"batch": 1, "channel": 3, "dimension": [224, 224]}],
    input_model_path="./examples/sample_models/graphmodule.pt",
    output_dir="./outputs/compressed/pytorch_automatic_compression_1",
    compression_ratio=0.5,
)