tensorflow / fold

Deep learning with dynamic computation graphs in TensorFlow
Apache License 2.0
1.82k stars 266 forks source link

How to use fold in serving? #86

Open KazuhiraDZ opened 6 years ago

KazuhiraDZ commented 6 years ago

My problem is, i want to implement a page for my members to test the model. I know tensorflow serving can finish it. But, is it support for fold? I run the code but it said 'Not found: Op type not registered 'DeserializingWeaver' in binary running on my pc' can the fold be able to support serving?

delesley commented 6 years ago

In theory I'm sure it's possible, but this is not something I've tried to do, so I don't have any instructions. You'll have to build from source, and it will probably take some work manipulating BUILD rules to make it happen. DeserializingWeaver is a custom TensorFlow Op used by Fold; you'll need to link it against serving in some way.

On Tue, Oct 10, 2017 at 7:50 PM, KazuhiraDZ notifications@github.com wrote:

My problem is, i want to implement a page for my members to test the model. I know tensorflow serving can finish it. But, is it support for fold? I run the code but it said 'Not found: Op type not registered 'DeserializingWeaver' in binary running on my pc' can the fold be able to support serving?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/tensorflow/fold/issues/86, or mute the thread https://github.com/notifications/unsubscribe-auth/AGGbTSuV8jVmHMnEhMFdYvtjpQXuoTKkks5srC0SgaJpZM4P016o .

-- DeLesley Hutchins | Software Engineer | delesley@google.com | 505-206-0315

KazuhiraDZ commented 6 years ago

@delesley Thank you for your advice. I have another question. Thanks to your Fold, I have train a model with large number of text trees. But it seems that I can't let it run with mutiGPUs even the code is correct in Tensorflow. I am anxious to train the model... so I want to know if the Fold can do.