access2rohit commented on a change in pull request #19025:
URL: https://github.com/apache/incubator-mxnet/pull/19025#discussion_r478701864
##########
File path: tests/nightly/test_np_large_array.py
##########
@@ -696,16 +696,16 @@ def test_one_hot():
assert A.grad[0] == 0
@use_np
[email protected](reason='backward value broken on large tensor')
def test_pick():
A = np.zeros((INT_OVERFLOW, 2))
B = np.zeros((INT_OVERFLOW))
+ A[0, 0] = 3
Review comment:
Since the operator is `pick` and it returns values of indices provided
in second argument. I wouldn't say its fixed then because you are unable to
access large indices by passing them as argument. You should perhaps look into
`operator=` implementation of this. Let me know if you think thats not that
case or if i could explain it a bit more clearly
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
