[MRG+1] FIX support memmap scalars as CV scores #6789
Conversation
| scoring=lambda est, X, y: scores) | ||
| except Exception: | ||
| raise | ||
| else: |
There was a problem hiding this comment.
Souldn't this be a 'finally', rather than an else?
There was a problem hiding this comment.
yes, that's what I mean
There was a problem hiding this comment.
Just not thinking straight today.
|
LGTM. +1 for merge |
| tf.write('Hello world!!!!!') | ||
| tf.close() | ||
| scores = np.memmap(tf.name, dtype=float) | ||
| score = scores.sum() |
There was a problem hiding this comment.
If I am following correctly your test relies on a sum of memmap to be a memmap 0d array. This is no longer the case in numpy master, see numpy/numpy#7406 for more details.
There was a problem hiding this comment.
You can use
score = np.memmap(tf.name, shape=(), mode=w+, dtype=float)`which will create a memmap 0d array. This will get rid of the Python 3.5 failure as well (you forgot the b prefix when writing to the file).
There was a problem hiding this comment.
Btw, I was hoping that was not permanent numpy behaviour, but it is because of that behaviour that this is an issue for us.
|
Thanks! |
* FIX support memmap scalars as CV scores * FIX test for Python 3.5 and NumPy 1.12
* FIX support memmap scalars as CV scores * FIX test for Python 3.5 and NumPy 1.12
Reference Issue
Fixes #6783
What does this implement/fix? Explain your changes.
The test - which introduces a cross-validation score of type np.memmap - fails without the patch.