corrected std function #767
Conversation
Codecov Report
@@ Coverage Diff @@
## master #767 +/- ##
==========================================
- Coverage 95.22% 95.22% -0.01%
==========================================
Files 64 64
Lines 8717 8728 +11
==========================================
+ Hits 8301 8311 +10
- Misses 416 417 +1
Flags with carried forward coverage won't be shown. Click here to find out more.
Continue to review full report at Codecov.
|
heat/core/statistics.py
Outdated
| if not isinstance(ddof, int): | ||
| raise TypeError(f"ddof must be integer, is {type(ddof)}") | ||
| elif ddof > 1: | ||
| raise NotImplementedError("Not implemented for ddof > 1.") |
There was a problem hiding this comment.
this is "only" not implemented for the distributed or GPU case, correct?
There was a problem hiding this comment.
i think it wasnt allowed by torch when i wrote that. but i guess it is now. ill take that out. worst case, we put it back in later
heat/core/statistics.py
Outdated
| unbiased = bool(ddof) | ||
| ddof = 1 if unbiased else ddof | ||
| if not x.is_distributed() and str(x.device)[:3] == "cpu": | ||
| loc = np.std(x._DNDarray__array.numpy(), axis=axis, ddof=ddof) |
There was a problem hiding this comment.
here ddof > 1 would be ok
There was a problem hiding this comment.
oh this was because numpy's std function is/was faster than torch's
| if not axis: | ||
| return np.sqrt(var(x, axis, ddof, **kwargs)) | ||
| if not isinstance(ddof, int): | ||
| raise TypeError(f"ddof must be integer, is {type(ddof)}") |
There was a problem hiding this comment.
This one line is not covered by tests.
Description
see title
Issue/s resolved: #766
Changes proposed:
Type of change
Due Diligence
Does this change modify the behaviour of other functions? If so, which?
no