qnq 1.1.2

Description:

qnq 1.1.2

QNQ -- QNQ's not quantization
version 1.1.0 2021.2.5
Description
The toolkit is for Techart algorithm team to quantize their custom neural network's pretrained model.
The toolkit is beta now, you can contact me with email([email protected]) for adding ops and fixing bugs.
How to install
pip install qnq
How to use
This README.MD is in very early stages, and will be updated soon.
you can visit https://git.zwdong.com/zhiwei.dong/qnq_tutorial for more examples for QNQ.


Prepare your model.

Check if your model contains non-class operator, like torch.matmul.
If True, add from qnq.operators.torchfunc_ops import * to your code.
Then use class replace non-class operator, you can refer fellow #! add by dongz

class BasicBlock(nn.Module):
expansion = 1

def __init__(self, inplanes, planes, stride=1, downsample=None):
super(BasicBlock, self).__init__()
self.conv1 = conv3x3(inplanes, planes, stride)
self.bn1 = nn.BatchNorm2d(planes)
self.relu1 = nn.ReLU(inplace=True)
self.relu2 = nn.ReLU(inplace=True)
self.conv2 = conv3x3(planes, planes)
self.bn2 = nn.BatchNorm2d(planes)
self.downsample = downsample
self.stride = stride

#! add by dongz
self.torch_add = TorchAdd()

def forward(self, x):
identity = x

out = self.conv1(x)
out = self.bn1(out)
out = self.relu1(out)

out = self.conv2(out)
out = self.bn2(out)

if self.downsample is not None:
identity = self.downsample(x)

#! add by dongz
out = self.torch_add(out, identity)
# out += identity
out = self.relu2(out)

return out



Prepare 'metrics', 'metrics_light'(optional) and 'steper'.

Choose at least 1k data to calibration your quantized model.
'metrics' inference without input params, return metrics value(a float number).
'metrics_light' inference without input params, return metrics value(a float number), you can choose 1/10 testsets to test.
'steper' done inference and without input params too, but add quant.step(), and no return.
Check qnq_tutorial for details.



Prepare pretrained checkpoints.

Train your model and use torch.save() to save your checkpoints.
Use checkpoints = torch.load(checkpoints_path) and model.load_state_dict(checkpoints) to load your checkpoints.



Quantize

For code

Add from qnq import QNQ
Add quant = QNQ(model, save_path, config_path, metrics, metrics_light, steper).
Add quant.search()


First run the program will exit, but the config_path will show a yaml file.
Edit config.yaml and rerun for quantization.



Operators supported

Convolution Layers

Conv
ConvTranspose


Pooling Layers

MaxPool
AveragePool
AdaptiveAvgPool


Activation

Relu、Relu6
PRelu、LeakyRelu
LogSoftmax


Normalization Layers

BatchNorm
LayerNorm


Recurrent

LSTM


Linear Layers

Linear


Vision Layers

Upsample
Embedding


Torch Function

Add, Sum, Minus, DotMul, MatMul, Div,
Sqrt, Exp
Sin, Cos
SoftMax, Sigmoid, Tanh
TorchTemplate, TorchDummy