This module implements versions of the key nn modules Conv2d() and Linear() which run in FP32 but with rounding applied to simulate the effect of INT8 quantization.
.. automodule:: torch.nn.qat
.. autoclass:: Conv2d :members:
.. autoclass:: Linear :members: