You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@mxnet.apache.org by GitBox <gi...@apache.org> on 2020/12/11 06:04:56 UTC

[GitHub] [incubator-mxnet] sxjscience opened a new issue #19659: [Bug] Latest MXNet fails the GluonNLP test case

sxjscience opened a new issue #19659:
URL: https://github.com/apache/incubator-mxnet/issues/19659


   ## Description
   Minimal reproducible example:
   
   @waytrue17 I think it might be related to the recent bugfix: https://github.com/apache/incubator-mxnet/pull/19614
   
   ```python
   import mxnet as mx
   from mxnet import nd, np, npx
   from mxnet.test_utils import assert_allclose
   from mxnet.gluon import HybridBlock, Constant
   from mxnet.gluon.data import DataLoader
   import itertools
   mx.npx.set_np()
   
   class Foo(HybridBlock):
       def forward(self, data, indices):
           mask = indices < 3
           data = npx.reshape(data, (-1, -2), reverse=True)
           mask = np.reshape(mask, (-1,))
           sel = nd.np._internal.boolean_mask(data, mask)
           return sel
   data = mx.np.random.normal(0, 1, (5, 5, 5, 5, 16))
   indices = mx.np.random.randint(0, 5, (5, 5, 5, 5))
   data.attach_grad()
   indices.attach_grad()
   foo = Foo()
   foo.hybridize()
   with mx.autograd.record():
       out = foo(data, indices)
       out.backward()
   out.wait_to_read()
   
   ```
   
   Output:
   
   ```
   [06:01:25] ../src/storage/storage.cc:199: Using Pooled (Naive) StorageManager for CPU
   
   [06:01:26] ../src/base.cc:80: cuDNN lib mismatch: linked-against version 8005 != compiled-against version 8004.  Set MXNET_CUDNN_LIB_CHECKING=0 to quiet this warning.
   ---------------------------------------------------------------------------
   MXNetError                                Traceback (most recent call last)
   <ipython-input-2-89d25fd8d10f> in <module>
        14 with mx.autograd.record():
        15     out = foo(data, indices)
   ---> 16     out.backward()
        17 out.wait_to_read()
   
   ~/env/nlp/lib/python3.6/site-packages/mxnet/ndarray/ndarray.py in backward(self, out_grad, retain_graph, train_mode)
      2908             ctypes.c_int(train_mode),
      2909             ctypes.c_void_p(0),
   -> 2910             ctypes.c_void_p(0)))
      2911 
      2912     def tostype(self, stype):
   
   ~/env/nlp/lib/python3.6/site-packages/mxnet/base.py in check_call(ret)
       244     """
       245     if ret != 0:
   --> 246         raise get_last_ffi_error()
       247 
       248 
   
   MXNetError: Traceback (most recent call last):
     File "../src/imperative/cached_op.cc", line 1150
   MXNetError: Check failed: s.op->num_backward_inputs() > bwd_in_dep + bwd_out_dep:
   ```


----------------------------------------------------------------
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:
users@infra.apache.org



---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@mxnet.apache.org
For additional commands, e-mail: issues-help@mxnet.apache.org