Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    Xipo
    @xipo
    Hello all,
    I'm getting ERR: Invalid GraphDef when trying to load a model that's in the SavedModel format
    Is the SavedModel format supported?
    Guy Korland
    @gkorland
    where did you get the model from ?
    did you train it or did you get it from the examples?
    Luca Antiga
    @lantiga_twitter
    Hi @xipo, RedisAI supports GraphDef (also known as Frozen Graph). SavedModel is a superset of that, see https://www.tensorflow.org/guide/extend/model_files, we don't support it at the moment.
    You can create a Frozen Graph easily using the converter package we have made - ml2rt.save_tensorflow
    https://github.com/hhsecond/ml2rt
    /cc @hhsecond
    Xipo
    @xipo
    @gkorland I trained it using an adapted tfx pipeline and now would like to test RedisAI as a replacement of tfx.serving
    @lantiga_twitter I had a look at that but didnt realize I can load a SavedModel
    Luca Antiga
    @lantiga_twitter
    @xipo Specifically exporting a frozen graph see https://github.com/hhsecond/ml2rt/blob/master/ml2rt/exporter.py#L8
    Xipo
    @xipo
    also thank you both for the info
    managed to convert to a GraphDef just now
    Luca Antiga
    @lantiga_twitter
    👍
    Sam Partee
    @Spartee
    Hello! Is there a mechanism for loading models onto cluster nodes like DAGRUN does with input tensors? This would be a way to ensure that if you have a cluster of GPU nodes, you could perform a modelrun and have it execute on any node of the cluster. e.g. have a copy of a model on each cluster node.