Use torch.amax in ReduceMax forward to support multiplie dim

[ice-tong]

Motivation

• We can use torch.amax in ReduceMax forward to support multiplie dim.
• The detail of torch.amax can be found at: https://pytorch.org/docs/stable/generated/torch.amax.html#torch-amax

  The difference between max/min and amax/amin is:

  • amax/amin supports reducing on multiple dimensions,
  • amax/amin does not return indices,
  • amax/amin evenly distributes gradient between equal values, while max(dim)/min(dim) propagates gradient only to a single index in the source tensor.

Modification

• ppq/executor/op/torch/default.py
  • Use torch.amax in ReduceMax forward to support multiplie dim