Getting AttributeError: 'NoneType' object has no attribute 'device_name' when trying to run the realistic example
sug4ndh opened this issue · comments
Hi,
I am trying to replicate the example as given here but I get AttributeError: 'NoneType' object has no attribute 'device_name' when I execute training_alice.py. I am using tensorflow version 1.13.2 as in the example and tf-encrypted version 0.5.9.
The full trace is as follows:
Traceback (most recent call last): File "training_alice.py", line 81, in <module> start_master("config.json") File "training_alice.py", line 77, in start_master main(server0) File "training_alice.py", line 49, in main model = LogisticRegression(num_features) File "/home/sug/ppml-work/common.py", line 12, in __init__ self.w_masked = tfe.mask(self.w) File "/home/sug/ppml/lib/python3.6/site-packages/tf_encrypted/protocol/pond/pond.py", line 938, in mask x_masked = _mask_private(self, x) File "/home/sug/ppml/lib/python3.6/site-packages/tf_encrypted/protocol/pond/pond.py", line 4035, in _mask_private a, a0, a1 = prot.triple_source.mask(x.backing_dtype, x.shape) File "/home/sug/ppml/lib/python3.6/site-packages/tf_encrypted/protocol/pond/triple_sources.py", line 47, in mask with tf.device(self.producer.device_name): AttributeError: 'NoneType' object has no attribute 'device_name'
Could you tell me what is causing the issue and how to rectify it?
Thanks!