Oracle offers GraphPipe spec for machine learning data transmission

(Foto: Istock)

Oracle offers GraphPipe spec for machine learning data transmission

IDG NEWS: GraphPipe is intended to bring the efficiency of a binary, memory-mapped format while being simple and light on dependencies

Vil du fortsette å lese, velg et av alternativene nedenfor

  • Logg inn!

    Du har abonnement og er registrert som bruker.

  • Har abonnement!

    Du har abonnement, men ikke registrert deg.

  • Bestill abonnement!

    Digital tilgang er inkludert i alle våre abonnement.

Oracle has developed an open source specification for transmitting tensor data, which the company wants to become a standard for machine learning.

also a intended querying Called being binary, memory-mapped servers dependencies. from format for and framework. efficiency any protocol the bring provides clients learning light is are while specification data GraphPipe, the and models There simple for GraphPipe network and a transmission. machine deploying of on to

includes: It

and are an Flatbuffer that a Flatbuffers during names. message additional protocol provide input flatbuffer - of names, similar includes A deserialization. tensors, input, memory buffers, to benefit request copy set Google avoiding with of definitions definitions. output

models. Guidelines serving for -

various models from serving learning machine Examples - frameworks. of

for plugin Python, for a model - Java. models Client and inside GraphPipe. There’s through available for TensorFlow remote for library, TensorFlow including Clients querying graph. a Go, served Google’s libraries local a are

and one remote and and metadata inputs outputs.  tensor accepts output model about provides also per a GraphPipe, of request shapes types With a name. message model returns The

Oracle … says GraphPipe

IDG News Service