capeprivacy / tf-trusted

tf-trusted allows you to run TensorFlow models in secure enclaves
https://capeprivacy.com/
Apache License 2.0
87 stars 11 forks source link

Is there any example provided for testing #24

Open HydraZeng opened 4 years ago

HydraZeng commented 4 years ago

Hi,

I am trying to test tf-trusted with some simple cases, but I don't know how to start. I find a logistic regression from tf-encrypted. I want to test more cases, like mnist with cnn. However, I don't know how to fill the arguments, e.g., input_name and output_name.

Any suggestion?

justin1121 commented 4 years ago

Hey there @HydraZeng,

In the README.md, there is a section about how to run a model found here. We don't have any better examples currently but if you're able to send your protobuf model file I could give you a hand choosing the right values for the input name and output name.

HydraZeng commented 4 years ago

Thanks very much for your response.

If I want to run the mnist_model.pb in here, and the input is in here. What are the input name and output name.

HydraZeng commented 4 years ago

I learn that it can be retrieved from netron, and I succeed in running the model I provide above. Thanks a lot

aanand300 commented 2 years ago

I am trying to run the same model and input mentioned in the above comment. After analyzing mnist_model.pb from netron browser, I am setting input_name as 'Placeholder' and output_name as 'fc2/add'. Can anyone confirm whether this is correct?