roywei commented on issue #8337: mx.autograd.grad works or fails depending on use of slices URL: https://github.com/apache/incubator-mxnet/issues/8337#issuecomment-379423074 @kpot may I know what's the purpose of slice? let's say if you want a[0:2], how you want to use it? `a[0:4]` is a shallow copy of `a` , using slice in autograd may not be the idea way. In your case for OpenCL device, could you try to use symbolic gradient directly from nnvm? refer to [gradient](https://github.com/dmlc/nnvm/blob/master/src/pass/gradient.cc) and [test](https://github.com/dmlc/nnvm/blob/master/tests/python/compiler/test_optimizer.py#L33) ``` a = mx.nd.array([1, 2, 3, 4], ctx=ctx) print(id(a)) print(id(a[0:4])) >>>4387046120 >>>4387048864
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services