def test_qat_mobile_per_channel_ff():
network = mobilenetV2(num_classes=1000)
img = Tensor(np.ones((1, 3, 224, 224)).astype(np.float32))
network = qat.convert_quant_network(network, bn_fold=True, per_channel=[False, False], symmetric=[True, False])
# should load the checkpoint. mock here
network.init_parameters_data()
qat.export(network, img, file_name="quant.pb")
def test_qat_mobile_per_channel_tf():
network = mobilenetV2(num_classes=1000)
img = Tensor(np.ones((1, 3, 224, 224)).astype(np.float32))
quantizer = QuantizationAwareTraining(bn_fold=True,
per_channel=[True, False],
symmetric=[True, False])
network = quantizer.quantize(network)
# should load the checkpoint. mock here
network.init_parameters_data()
quant_export.export(network, img, file_name="quant.pb")
def test_lsq_mobile_per_channel_ff():
network = mobilenetV2(num_classes=1000)
img = Tensor(np.ones((1, 3, 224, 224)).astype(np.float32))
quantizer = QuantizationAwareTraining(
bn_fold=True,
per_channel=[False, False],
symmetric=[True, True],
narrow_range=[True, True],
freeze_bn=0,
quant_delay=0,
one_conv_fold=True,
optimize_option=OptimizeOption.LEARNED_SCALE)
network = quantizer.quantize(network)
# should load the checkpoint. mock here
network.init_parameters_data()
quant_export.export(network, img, file_name="quant.pb")