Tanh does not work under gpu

TensorFlow General Discussion
,
1

I’m trying to evaluate the hyperbolic tangent of a Tensorflow vector. The operation works fine under TensorFlow but tensorflow-gpu provides the following error:

 I_so = (0.5 * (a_so - tau_a) * (1 + tf.tanh((U - b_so) / c_so)) +
  File "/home/cc14/Software/anaconda3/envs/tf-gpu/lib/python3.8/site-packages/tensorflow/python/ops/gen_math_ops.py", line 10529, in tanh
    _ops.raise_from_not_ok_status(e, name)
  File "/home/cc14/Software/anaconda3/envs/tf-gpu/lib/python3.8/site-packages/tensorflow/python/framework/ops.py", line 6862, in raise_from_not_ok_status
    six.raise_from(core._status_to_exception(e.code, message), None)
  File "<string>", line 3, in raise_from
tensorflow.python.framework.errors_impl.InternalError: Failed to load in-memory CUBIN: CUDA_ERROR_NO_BINARY_FOR_GPU: no kernel image is available for execution on the device [Op:Tanh]

what did generate it? Is there a way to use tf.tanh under GPU?

thanks,

Cesare

3

Hi Cesare!

I’m able to use tf.tanh on the GPU, see below. Here are two links that have helped me in the past with these sorts of issues:

Installation using pip
GPU guide

Pritam

[ins] In [6]: tf.config.list_physical_devices('GPU')                                                                                                                                                       
Out[6]: [PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]                                                                                                                                 
                                                                                                                                                                                                           
[nav] In [7]: x = tf.constant([-float("inf"), -5, -0.5, 1, 1.2, 2, 3, float("inf")])                                                                                                                       
                                                                                                                                                                                                           
[nav] In [8]: tf.tanh(x)                                                                                                                                                                                   
Out[8]:                                                                                                                                                                                                    
<tf.Tensor: shape=(8,), dtype=float32, numpy=                                                                                                                                                              
array([-1.        , -0.99990916, -0.46211717,  0.7615942 ,  0.8336546 ,                                                                                                                                    
        0.9640276 ,  0.9950547 ,  1.        ], dtype=float32)>