I am trying to do some image batch predictions on a custom model prediction routine. I have followed this tutorial for creating the routine wrapper. I have deployed my model as a tar.gz on google cloud ai platform, created a version and so on. Now when I try to launch a batch prediction from command line using this command:
gcloud ai-platform jobs submit prediction $JOB_NAME --model $MODEL_NAME \
--input-paths $INPUT_PATHS \
--output-path $OUTPUT_PATH \
--region $REGION \
--data-format $DATA_FORMAT
I get the following error
ERROR: (gcloud.ai-platform.jobs.submit.prediction) INVALID_ARGUMENT: Field: prediction_input.framework Error: Batch prediction only supports Tensorflow.FRAMEWORK_CUSTOM_CLASS
- '@type': type.googleapis.com/google.rpc.BadRequest
fieldViolations:
- description: Batch prediction only supports Tensorflow.FRAMEWORK_CUSTOM_CLASS
field: prediction_input.framework
The input is in JSON format and I think it is structured in the right way, because in the bucket, on the console, under type it says {"instances": [{"image_bytes": {"b64": "image/jpeg"}}]}
, and if I click on the JSON it show straight the image that it contains. Pretty cool!
But I don't understand the error.