Description:
When ComputationModel._collective_op_dtype is None and the input to a
collective op is a float number, the constructed intermediate tensor is of
dtype torch.float32 rather than torch.double, hence we would have a loss
in precision.

Check list:

• New tests are added (if a new feature is added)
• New doc strings: description and/or example code are in RST format
• Documentation is updated (if required)