Thank you for your post. We noticed you have not filled out the following field in the issue template. Could you update them if they are relevant in your case, or leave them as N/A? Thanks.
Have I written custom code
OS Platform and Distribution
TensorFlow installed from
TensorFlow version
Bazel version
CUDA/cuDNN version
GPU model and memory
Exact command to reproduce

/CC @dsmilkov, can you take a look at this proposal?

Hi @cancan101, the body of the py_func will not be serialized in a GraphDef. Therefore, you should not use this function if you need to serialize your model and restore it in a different environment.

Also, if you are using TensorFlow.js for development and thinking about this in a JS-only context, would you mind re-posting this issue to the tensorflow/tfjs issues?

Thanks!

Interesting to know that the body is not serialized; I had not realized that. I was not thinking of js_func in the context of tfjs. Rather I was thinking of loading a graph on iOS or Android and using the js_func to supply user created functions.

/CC @andrewharp can you take a look since this is desired in mobile?

@cancan101,
Please let me know if my understanding is correct here. Looks like this isssue is more related to TensorFlow JavaScript, if you are using TensorFlow.js for development and thinking about this in a Javascript context, could you please try to post this issue to the tensorflow/tfjs issues

Thank you!

No, this is not a tfjs issue. This is about the ability to have a node that can run javascript, similar to the py_func node that runs python code.