quark.torch.quantization.graph.optimization.replace_conv2d_to_qtconv2d

Module Contents

Functions

quark.torch.quantization.graph.optimization.replace_conv2d_to_qtconv2d.replace_conv2d_qtconv2d(m: torch.fx.GraphModule) → None

replace [ops.aten.conv2d] to QuantConv2d ops.aten.conv2d:

args: (Tensor input, Tensor weight, Tensor? bias=None, SymInt[2] stride=1, SymInt[2] padding=0, SymInt[2] dilation=1, SymInt groups=1) -> Tensor required: [input, weight] optional: [bias=None, SymInt[2] stride=1, SymInt[2] padding=0, SymInt[2] dilation=1, SymInt groups=1]