0

I am using google vertex AI online prediction: In order to send an image it has to be in a JSON file in unit8 format which has to be less than 1.5 MB, when converting my image to uint8 it definitely exceeds 1.5MB.

To go around this issue we can encode the unit8 file to b64, that makes the JSON file in KBs when running the prediction I get Resource Exhausted: 429 received trailing metadata size exceeds limit Is there anyone who knows what's the problem?

ARAFAT
  • 39
  • 3
  • Have you followed any of these tutorials from the documentation: [Hello Image data](https://cloud.google.com/vertex-ai/docs/tutorials/image-recognition-automl#before_you_begin) or [Hello custom training](https://cloud.google.com/vertex-ai/docs/tutorials/image-recognition-custom#before_you_begin)? Both tutorials explain how to train and deploy a model to analyse and predict flower types from the images. – Alexandre Moraes Jul 07 '21 at 09:12
  • @AlexandreMoraes, I don't think you understand my problem, I already have gone through all the docs. In my model batch prediction works fine but this error happens when I request an online prediction. – ARAFAT Jul 07 '21 at 10:50
  • I understand, the API expects a JSON input for online prediction, have you prepared your data as explained here: [link1](https://cloud.google.com/vertex-ai/docs/predictions/online-predictions-custom-models) or [link2](https://medium.com/devopsturkiye/how-to-operate-a-model-on-vertex-ai-from-google-cloud-888c9dd3e7bd)? – Alexandre Moraes Jul 14 '21 at 06:37
  • Yes............ – ARAFAT Jul 14 '21 at 07:10
  • As seen from the comments the answer does not show what is attempted but only describes it. More detail is needed so I would vote to close it as it can probably not be answered in it's current form – Dennis Jaheruddin Apr 07 '23 at 21:01

0 Answers0