QAT¶

class paddle.quantization. QAT ( config: paddle.quantization.config.QuantConfig ) [source]

Tools used to prepare model for quantization-aware training. :param config: :type config: QuantConfig

Examples

System Message: ERROR/3 (/usr/local/lib/python3.8/site-packages/paddle/quantization/qat.py:docstring of paddle.quantization.qat.QAT, line 7)

Error in “code-block” directive: maximum 1 argument(s) allowed, 20 supplied.

.. code-block:: python
    from paddle.quantization import QAT, QuantConfig
    from paddle.quantization.quanters import FakeQuanterWithAbsMaxObserver
    quanter = FakeQuanterWithAbsMaxObserver(moving_rate=0.9)
    q_config = QuantConfig(activation=quanter, weight=quanter)
    qat = QAT(q_config)

quantize ( model: paddle.nn.layer.layers.Layer, inplace=False ) [source]

quantize¶

Create a model for quantization-aware training.

The quantization configuration will be propagated in the model. And it will insert fake quanters into the model to simulate the quantization.

Parameters

• model (Layer) –

• inplace (bool) –

Return: The prepared model for quantization-aware training.

Examples: .. code-block:: python

System Message: ERROR/3 (/usr/local/lib/python3.8/site-packages/paddle/quantization/qat.py:docstring of paddle.quantization.qat.QAT.quantize, line 15)

Unexpected indentation.

from paddle.quantization import QAT, QuantConfig from paddle.quantization.quanters import FakeQuanterWithAbsMaxObserver from paddle.vision.models import LeNet

quanter = FakeQuanterWithAbsMaxObserver(moving_rate=0.9) q_config = QuantConfig(activation=quanter, weight=quanter) qat = QAT(q_config) model = LeNet() quant_model = qat.quantize(model) print(quant_model)

convert ( model: paddle.nn.layer.layers.Layer, inplace=False )

convert¶

Convert the quantization model to onnx style. And the converted model can be saved as inference model by calling paddle.jit.save. :param model: :type model: Layer :param inplace: :type inplace: bool

Return: The converted model

Examples: .. code-block:: python

System Message: ERROR/3 (/usr/local/lib/python3.8/site-packages/paddle/quantization/qat.py:docstring of paddle.quantization.quantize.Quantization.convert, line 12)

Unexpected indentation.

import paddle from paddle.quantization import QAT, QuantConfig from paddle.quantization.quanters import FakeQuanterWithAbsMaxObserver from paddle.vision.models import LeNet

quanter = FakeQuanterWithAbsMaxObserver(moving_rate=0.9) q_config = QuantConfig(activation=quanter, weight=quanter) qat = QAT(q_config) model = LeNet() quantized_model = qat.quantize(model) converted_model = qat.convert(quantized_model) dummy_data = paddle.rand([1, 1, 32, 32], dtype=”float32”) paddle.jit.save(converted_model, “./quant_deploy”, [dummy_data])