search

NervanaSystems / he-transformer

nGraph-HE: Deep learning with Homomorphic Encryption (HE) through Intel nGraph
Apache License 2.0
221 stars 39 forks source link

Error when test ax.py #159

Open allensll opened 5 years ago

allensll commented 5 years ago

Test ax.py, and it stop, not wait for a client to connect.

(venv-tf-py3) allensll@allensll:~/nGraph-HE/he-transformer/examples$ NGRAPH_ENABLE_CLIENT=1 NGRAPH_ENCRYPT_DATA=1 NGRAPH_TF_BACKEND=HE_SEAL_CKKS python ax.py
2019-04-13 21:34:33.838920: I tensorflow/core/platform/profile_utils/cpu_utils.cc:94] CPU Frequency: 3192000000 Hz
2019-04-13 21:34:33.839354: I tensorflow/compiler/xla/service/service.cc:150] XLA service 0x55564b6eaa80 executing computations on platform Host. Devices:
2019-04-13 21:34:33.839399: I tensorflow/compiler/xla/service/service.cc:158]   StreamExecutor device (0): <undefined>, <undefined>
Result:  [[ 2.  6.]
 [12. 20.]]
(venv-tf-py3) allensll@allensll:~/nGraph-HE/he-transformer/examples$

Thanks.

fboemer commented 5 years ago

I believe TensorFlow simplifies simple expressions automatically. You could try more complex expressions rather than (a+b)*a*b. Maybe try ReLU or MaxPool, or square.

allensll commented 5 years ago

I believe TensorFlow simplifies simple expressions automatically. You could try more complex expressions rather than (a+b)*a*b. Maybe try ReLU or MaxPool, or square.

I tried the MNIST-MLP and MNIST-Cryptonets, and the same error occurred.

AlexMV12 commented 5 years ago

I have the same problem, it's like NGRAPH_ENABLE_CLIENT=1 is ignored: ax.py continues as normal, computing the small example without waiting for a client.