kube-HPC / hkube

🐟 High Performance Computing over Kubernetes - Core Repo 🎣
http://hkube.io
MIT License
305 stars 20 forks source link

differentiate batch pipeline wizard from streaming #1845

Closed golanha closed 6 months ago

golanha commented 8 months ago

We'd like to direct the user to use the appropriate pipeline wizard parts according to his aim, streaming or batch.

  1. When streaming pipeline is selected: a. When adding a new node, kind will include the stateless/statefull instead of algotihm/datasource/... b. A new input for a node will only contain a text box, no drop down, it will be treated as value input. c. In node options no need for "include in pipeline result" and "Create a tensorboard"

  2. When toggeling between streaming and batch pipeline, warn. a. From batch to streaming, warn of loosing any input of type outputof, outputof multy, multi value. If user selected other kind than algorithm also warn.

  3. By converting the pipeline to a Streaming you will lose some of the input parameters you defined.

  4. There is no use of <datasource/hyerparam/...> in streaming, <node/nodes> will be removed.

Do you wish to precede?


 b. From streaming to batch, warn of loosing defined flows, and if statetype had been selected warn, this information including min max are about to be lost.
  1. By converting the pipeline to batch you will lose the streaming flows you defined.
  2. You set some of the nodes to stateless or statefull, there is no meaning for that in batch pipeline. You will lose any information related to these kinds. Do you wish to precede?