Hi, tracing a simple quantized model gives me a graph that looks like this:
graph(%self.1 : __torch__.___torch_mangle_2.MyCell,
%x : Float(10, 4, strides=[4, 1], requires_grad=0, device=cpu)):
%22 : __torch__.torch.nn.quantized.dynamic.modules.linear.Linear = prim::GetAttr[name="linear"](%self.1)
%25 : __torch__.torch.nn.quantized.modules.linear.LinearPackedParams = prim::GetAttr[name="_packed_params"](%22)
%26 : __torch__.torch.classes.quantized.LinearPackedParamsBase = prim::GetAttr[name="_packed_params"](%25)
%27 : bool = prim::Constant[value=1](), scope: __module.linear # /Users/kohenchia/development/pytorch/.virtualenv/lib/python3.8/site-packages/torch/nn/quantized/dynamic/modules/linear.py:47:0
%Y : Float(10, 7, strides=[7, 1], requires_grad=0, device=cpu) = quantized::linear_dynamic(%x, %26, %27), scope: __module.linear # /Users/kohenchia/development/pytorch/.virtualenv/lib/python3.8/site-packages/torch/nn/quantized/dynamic/modules/linear.py:47:0
%29 : int = prim::Constant[value=6](), scope: __module.linear # /Users/kohenchia/development/pytorch/.virtualenv/lib/python3.8/site-packages/torch/nn/quantized/dynamic/modules/linear.py:54:0
%30 : bool = prim::Constant[value=0](), scope: __module.linear # /Users/kohenchia/development/pytorch/.virtualenv/lib/python3.8/site-packages/torch/nn/quantized/dynamic/modules/linear.py:54:0
%31 : bool = prim::Constant[value=0](), scope: __module.linear # /Users/kohenchia/development/pytorch/.virtualenv/lib/python3.8/site-packages/torch/nn/quantized/dynamic/modules/linear.py:54:0
%32 : NoneType = prim::Constant(), scope: __module.linear
%33 : Float(10, 7, strides=[7, 1], requires_grad=0, device=cpu) = aten::to(%Y, %29, %30, %31, %32), scope: __module.linear # /Users/kohenchia/development/pytorch/.virtualenv/lib/python3.8/site-packages/torch/nn/quantized/dynamic/modules/linear.py:54:0
return (%33)
How do I inspect the quantization parameters for the weights in this model? The weights are statically quantized.