Intel MKL ERROR: Parameter 4 was incorrect on entry to DLASCL.

在运行程序的时候，出现了以下错误提示；

Intel MKL ERROR: Parameter 4 was incorrect on entry to DLASCL.

Intel MKL ERROR: Parameter 4 was incorrect on entry to DLASCL.
Traceback (most recent call last):
  File "task2.py", line 322, in <module>
    sys.exit(main(sys.argv))
  File "task2.py", line 256, in main
    app.system_training()
  File "/disk2/zoe-dir/DCASE2017-baseline-system-1.2.0/dcase_framework/decorators.py", line 38, in function_wrapper
g
    data_filenames=data_filelist
  File "/disk2/zoe-dir/DCASE2017-baseline-system-1.2.0/dcase_framework/learners.py", line 2322, in learn
    self.create_model(input_shape=(80,40,1))  # input_shape=200
  File "/disk2/zoe-dir/DCASE2017-baseline-system-1.2.0/dcase_framework/keras_utils.py", line 380, in create_model
    self.model.add(GRU(96,activation='tanh',name='gru1'))  #,recurrent_activation='hard_sigmoid'
  File "/home/zoe/anaconda2/lib/python2.7/site-packages/keras/models.py", line 475, in add
    output_tensor = layer(self.outputs[0])
  File "/home/zoe/anaconda2/lib/python2.7/site-packages/keras/layers/recurrent.py", line 268, in __call__
    return super(Recurrent, self).__call__(inputs, **kwargs)
  File "/home/zoe/anaconda2/lib/python2.7/site-packages/keras/engine/topology.py", line 575, in __call__
    self.build(input_shapes[0])
  File "/home/zoe/anaconda2/lib/python2.7/site-packages/keras/layers/recurrent.py", line 754, in build
    constraint=self.recurrent_constraint)
  File "/home/zoe/anaconda2/lib/python2.7/site-packages/keras/legacy/interfaces.py", line 87, in wrapper
    return func(*args, **kwargs)
  File "/home/zoe/anaconda2/lib/python2.7/site-packages/keras/engine/topology.py", line 396, in add_weight
    weight = K.variable(initializer(shape),
  File "/home/zoe/anaconda2/lib/python2.7/site-packages/keras/initializers.py", line 247, in __call__
    u, _, v = np.linalg.svd(a, full_matrices=False)
  File "/home/zoe/anaconda2/lib/python2.7/site-packages/numpy/linalg/linalg.py", line 1389, in svd
    u, s, vt = gufunc(a, signature=signature, extobj=extobj)
  File "/home/zoe/anaconda2/lib/python2.7/site-packages/numpy/linalg/linalg.py", line 99, in _raise_linalgerror_svd_nonconvergence
    raise LinAlgError("SVD did not converge")
numpy.linalg.linalg.LinAlgError: SVD did not converge

问题出在包mkl（2018.0.0）中，似乎它最近已经发布并且与通过conda *提供的Tensorflow（1.3.0）和Keras（2.0.5）提供的一些软件包的版本冲突。

因此，我使用Anaconda Navigator手动降级mkl到v11.3.3，这导致其他软件包自动降级，现在一切正常。

解决方法：更新mkl：
conda install mkl==11.3.3
问题解决！

参考链接：
Keras with Tensorflow backend on GPU. MKL ERROR: Parameter 4 was incorrect on entry to DLASCL

最后编辑于：2018.07.24 10:45:43