gistfile1.txt

# script

import torch                                                     
import torch.nn as nn                                            
import torch.nn.quantized as nnq                                 
                                                                 
m = nn.Sequential(nn.Conv2d(1, 1, 1))                            
m.qconfig = torch.quantization.get_default_qat_qconfig('qnnpack')
mp = torch.quantization.prepare_qat(m)                           
print('prepared model', mp)                                      
mp(torch.randn(4, 1, 4, 4))                                      
mq = torch.quantization.convert(mp)                              
print('quantized model', mq)                                     
print('prepared parameters')                                     
print(list(mp.named_parameters()))                               
print('quantized packed parameters')                             
for name, mod in mq.named_modules():                             
    if isinstance(mod, nnq.Conv2d):                              
        weight, bias = mod._weight_bias()                        
        print(name, 'weight', weight, 'bias', bias)              
        

# output


prepared model Sequential(
  (0): Conv2d(
    1, 1, kernel_size=(1, 1), stride=(1, 1)
    (weight_fake_quant): FakeQuantize(
      fake_quant_enabled=tensor([1], dtype=torch.uint8), observer_enabled=tensor([1], dtype=torch.uint8), quant_m
in=-128, quant_max=127, dtype=torch.qint8, qscheme=torch.per_tensor_symmetric, ch_axis=-1, scale=tensor([1.]), ze
ro_point=tensor([0])
      (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
    )
    (activation_post_process): FakeQuantize(
      fake_quant_enabled=tensor([1], dtype=torch.uint8), observer_enabled=tensor([1], dtype=torch.uint8), quant_m
in=0, quant_max=255, dtype=torch.quint8, qscheme=torch.per_tensor_affine, ch_axis=-1, scale=tensor([1.]), zero_po
int=tensor([0])
      (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
    )
  )
)
quantized model Sequential(
  (0): QuantizedConv2d(1, 1, kernel_size=(1, 1), stride=(1, 1), scale=0.006431101355701685, zero_point=109)
)
prepared parameters
[('0.weight', Parameter containing:
tensor([[[[0.4274]]]], requires_grad=True)), ('0.bias', Parameter containing:
tensor([0.1753], requires_grad=True))]
quantized packed parameters
0 weight tensor([[[[0.4257]]]], size=(1, 1, 1, 1), dtype=torch.qint8,
       quantization_scheme=torch.per_tensor_affine, scale=0.0033521773293614388,
       zero_point=0) bias tensor([0.1753], requires_grad=True)