def deployable_model(self, src_dir, used_for_xmodel=False):
if used_for_xmodel:
device = torch.device('cpu')
inputs = self._inputs.to(device)
else:
device = self._device
inputs = self._inputs
model = copy.deepcopy(self._model)
model.load_state_dict(
torch.load(os.path.join(src_dir, _DEPLOYABLE_MODEL_NAME)))
qprocessor = qproc.TorchQuantProcessor(
'test',
model,
inputs,
output_dir=src_dir,
bitwidth_w=self._bitwidth,
bitwidth_a=self._bitwidth,
mix_bit=self._mix_bit,
device=device)
self._qprocessor = qprocessor
if used_for_xmodel:
logging.info(
'Forward the deployable model with data of batch_size=1 in cpu mode to dump xmodel.'
)
return qprocessor.quant_model()
def fuse_conv_bn(model):
model.apply(conv_fused.fuse_conv_bn)
model.conv_bn_fused = True
logging.info('Merge batchnorm to conv.')
def freeze_bn_stats(model):
model.apply(conv_fused.freeze_bn_stats)
logging.info('Running statistics of batch normlization has been frozen.')
def freeze_quant(model):
model.apply(quantizer_mod.freeze_quant)
logging.info('Scale of quantizer has been frozen.')
def enable_warmup(model):
model.apply(quantizer_mod.enable_warmup)
logging.info('Initialize quantizer.')
def disable_quant(model):
model.apply(quantizer_mod.disable_quant)
logging.info(
'Disable quantization: floating point operations will be performed.')
def enable_quant(model):
model.apply(quantizer_mod.enable_quant)
logging.info('Enable quantization: quantized operations will be performed.')