- class torch.ao.nn.qat.Conv3d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros', qconfig=None, device=None, dtype=None)¶
A Conv3d module attached with FakeQuantize modules for weight, used for quantization aware training.
We adopt the same interface as torch.nn.Conv3d, please see https://pytorch.org/docs/stable/nn.html?highlight=conv3d#torch.nn.Conv3d for documentation.
Similar to torch.nn.Conv3d, with FakeQuantize modules initialized to default.
weight_fake_quant – fake quant module for weight