Erorr occured when I modify real-time inference to batch transform about huggingface-sagemaker-workshop-series

[philschmid]

You can find the code here: https://github.com/philschmid/huggingface-sagemaker-workshop-series/tree/main/workshop_3_mlops/utils

[philschmid]

Could you please format your code to read it better and also share the error you are seeing? And is that the complete code for your example?

[isoyaoya]

Have you tried batch transform inference rather than realtime inference before? Could you please provide some sample codes?

I have attached the whole notebook. Now there are some other errors:

lab_1_sagemaker_pipeline_batchtransform.ipynb.zip

[philschmid]

Could you please add the whole error stack it seems that there is something missing.

[isoyaoya]

log-events-viewer-result (2).csv

[isoyaoya]

Attach the whole error stack, thank you so much!

[philschmid]

From error it seems that you are sending malformed payloads. Here is an example for batch transform: https://github.com/huggingface/notebooks/blob/main/sagemaker/12_batch_transform_inference/sagemaker-notebook.ipynb

[isoyaoya]

I have seen this code before. Of course, if I do not use Sagemaker pipeline, I can create batch transform job without error and output correct predict results. However, I was required to use Sagemaker pipeline and batch transform to achieve automaticlly process.

[isoyaoya]

Have you tried to modify your code from real-time inference to batch transform with Sagemaker pipeline before? In some NLP applications, batch transform is often used.

[isoyaoya]

Use a lambda, create a lambda step to deploy registered model could solve the issue.

[dilky-ascentic]

@isoyaoya can you share the how you configured lambda step to do batch transform please?