Allow to specify model version for tensorflow serving

[zewelor]

Allow to specify model version as shown here https://www.tensorflow.org/tfx/serving/api_rest#predict_api . From looking at the code, currently it doesn't seem possible: https://github.com/SeldonIO/seldon-core/blob/master/integrations/tfserving/TfServingProxy.py#L45

[ukclivecox]

If you use 1.1.0 of Seldon you can use the Tensorflow protocol directly. https://docs.seldon.io/projects/seldon-core/en/latest/graph/protocols.html#rest-and-grpc-tensorflow-protocol

Its true we don't expose the version path in REST

seldon-core/executor/api/rest/server.go

Line 150 in 33d66ae

r.Router.NewRoute().Path("/v1/models/{" + ModelHttpPathVariable + "}/:predict").Methods("POST").HandlerFunc(r.wrapMetrics(metric.PredictionHttpServiceName, r.predictions))

In grpc you should be able to use it at present.

[zewelor]

Thanks I will look into it.

[ukclivecox]

Will close this assuming using Tensoflow protocol is the best solution as suggested. Please reopen if still an issue.