ConvBnRelu2d¶

class ConvBnRelu2d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, conv_mode='cross_correlation', compute_mode='default', eps=1e-05, momentum=0.9, affine=True, track_running_stats=True, **kwargs)[源代码]¶: A fused QATModule including Conv2d, BatchNorm2d and relu with QAT support. Could be applied with Observer and FakeQuantize.

ConvBn2d

BatchMatMulActivation