[GitHub] anirudh2290 commented on issue #8288: test_svmoutput_with_type fails in CI builds

[GitBox <gi...@apache.org>]

anirudh2290 commented on issue #8288: test_svmoutput_with_type fails in CI builds
URL: https://github.com/apache/incubator-mxnet/issues/8288#issuecomment-399620847
 
 
   Reproducible script :
   
   ```
   import mxnet as mx
   import numpy as np
   
   def try_svm(dtype, x):
       exe = sym.simple_bind(grad_req='write',**{'ctx': mx.cpu(0), 'svmoutput_data': (4, 2), 'type_dict': {'svmoutput_data': dtype} })
       exe.arg_arrays[0][:] = x.astype(dtype)
       exe.forward(is_train=True)
       exe.backward(exe.outputs)
       print(exe.arg_arrays[0].asnumpy())
       print(exe.grad_arrays[0].asnumpy())
   
   np.random.seed(1640822401)
   x = np.random.normal(size=(4, 2), scale=1.0)
   sym = mx.sym.SVMOutput(name='svmoutput', use_linear=True)
   try_svm(np.float32, x)
   try_svm(np.float16, x)
   ```
   
   You will find that the two grad_arrays are different.

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


With regards,
Apache Git Services