Hi, Yury, Nikolay. This issue reproduces in "TensorFlow on Apache Ignite" use cases. When user prepares training script (like official MNIST model https://github.com/tensorflow/models/tree/master/official/mnist), runs it in distributed standalone client mode (see this documentation https://github.com/tensorflow/tensorflow/tree/master/tensorflow/contrib/distribute#standalone-client-mode) using IgniteDataset he can see that workers throw exceptions says that IgniteDataset operation is not loaded.
For more information about distributed training please see https://docs.google.com/document/d/1PtYmyaT4dPBNp1mWkCczOdCZHt2xaqB183honfMup3g/edit documentation. -- Sent from: http://apache-ignite-developers.2346864.n4.nabble.com/
