0

When trying to deploy a keras model to ML Engine I get

$ gcloud ml-engine versions create v2 --model=plantDisease01 --origin=gs://${BUCKET_NAME}/
plantDisease01 --runtime-version=1.4                                                       
Creating version (this might take a few minutes)......failed.                              ERROR: (gcloud.ml-engine.versions.create) Bad model detected with error:  "Failed to load model: Loading servable: {name: default version: 1} failed: Not found: Op type not registere
d 'ClipByValue' in binary running on localhost. Make sure the Op and Kernel are registered
in the binary running in this process.\n\n (Error code: 0)"
FAIL

my storage looks like

$ gsutil ls gs://keras-class-191806/plantDisease01/export                      [23:29:38]
gs://keras-class-191806/plantDisease01/export/
gs://keras-class-191806/plantDisease01/export/saved_model.pb

I built the protocol buffer version using this approach https://stackoverflow.com/a/44232441/630752

Harry Moreno
  • 10,231
  • 7
  • 64
  • 116

1 Answers1

1

ClipByValue was introduced in TensorFlow 1.8. You can either register the op yourself or just change the --runtime-version flag's value to 1.8.

Lefteris S
  • 1,614
  • 1
  • 7
  • 14