Oracle, is playing in an unlikely terrain as the heavily locked-down company has open-sourced its tool created to make the deployment of machine learning models in the cloud easy, GraphPipe.

While popular frameworks like Google's TensorFlow and Amazon's Caffe2 are already leading in the machine learning deployment verticals, GraphPipe aims to make it easier to deploy machine learning models for use on mobile apps and IoT devices, and to serve for internal use within organizations.

According to Oracle, GraphPipe is created to solve three challenges: a standard, high-performance protocol for transmitting tensor data over the network, along with simple implementations of clients and servers that make deploying and querying machine learning models from any framework easy.

It can serve models built on TensorFlow, PyTorch, mxnet, CNTK, or caffe2; and developers don’t need custom APIs to deploy AI models or require popular framework to create a model.

The adoption of machine learning in the enterprise has been rather slower than expected, as organizations find it difficult to manage their own machine learning technology, and the models are often deployed using bespoke techniques, which is difficult to manage across servers in different departments.

This challenge is what Oracle is aiming to solve with the new open source, high-performance standard for transmitting tensor data.

GraphPipe is available on GitHub, with documentation, examples, and other relevant content available at this web address: https://oracle.github.io/graphpipe.

GraphPipe, Oracle's open-source standard for deploying Machine Learning models



Oracle, is playing in an unlikely terrain as the heavily locked-down company has open-sourced its tool created to make the deployment of machine learning models in the cloud easy, GraphPipe.

While popular frameworks like Google's TensorFlow and Amazon's Caffe2 are already leading in the machine learning deployment verticals, GraphPipe aims to make it easier to deploy machine learning models for use on mobile apps and IoT devices, and to serve for internal use within organizations.

According to Oracle, GraphPipe is created to solve three challenges: a standard, high-performance protocol for transmitting tensor data over the network, along with simple implementations of clients and servers that make deploying and querying machine learning models from any framework easy.

It can serve models built on TensorFlow, PyTorch, mxnet, CNTK, or caffe2; and developers don’t need custom APIs to deploy AI models or require popular framework to create a model.

The adoption of machine learning in the enterprise has been rather slower than expected, as organizations find it difficult to manage their own machine learning technology, and the models are often deployed using bespoke techniques, which is difficult to manage across servers in different departments.

This challenge is what Oracle is aiming to solve with the new open source, high-performance standard for transmitting tensor data.

GraphPipe is available on GitHub, with documentation, examples, and other relevant content available at this web address: https://oracle.github.io/graphpipe.

No comments