I built a GPU-based custom op for GELU operation, and would like to call the op within a python script. However, I have been getting the following error message as the op executes:
Traceback (most recent call last):
File “use_op.py”, line 4, in
gelu_module.gelu_op([2.0, 3.0])
File “”, line 47, in gelu_op
File “”, line 85, in gelu_op_eager_fallback
File “/usr/local/lib/python3.8/dist-packages/tensorflow/python/eager/execute.py”, line 59, in quick_execute
tensors = pywrap_tfe.TFE_Py_Execute(ctx._handle, device_name, op_name,
tensorflow.python.framework.errors_impl.NotFoundError: Could not find device for node: {{node GeluOp}} = GeluOp[T=DT_FLOAT]
All kernels registered for op GeluOp:
<no registered kernels>
[Op:GeluOp]
From the message, it seems that no op is registered. Below is a link to access source code, if someone is interested in helping me figure out the issue: remote repo. Thanks.