【cherry-pick】Setitem support passing stop_gradient from value to tensor (#37023)

[zyfncg]

PR types

Bug fixes

PR changes

OPs

Describe

Fix issue:36902

Before this PR:

a = paddle.rand(shape=[1,4])
b = paddle.rand(shape=[1,4])
a.stop_gradient = False
b.stop_gradient = False

d = paddle.zeros((4, 4))
print(d.stop_gradient)  # True
c = a/b
d[0, :] = a/b

print(c.stop_gradient)  # False
print('Is d requires grad: ', not d.stop_gradient)  # Is d requires grad:  False

Now:

a = paddle.rand(shape=[1,4])
b = paddle.rand(shape=[1,4])
a.stop_gradient = False
b.stop_gradient = False

d = paddle.zeros((4, 4))
print(d.stop_gradient)  # True
c = a/b
d[0, :] = a/b

print(c.stop_gradient)  # False
print('Is d requires grad: ', not d.stop_gradient)  # Is d requires grad:  True

[paddle-bot-old]

Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

[MingMingShangTian]

LGTM