Error in Sagemaker clarify processing job for the explainability monitoring job

It's hard to tell exactly what is causing the error without a stacktrace, or more information on the inputs. That VisibleDeprecationWarning may not necessarily be where the error is occurring, but you may be able to tell from the stacktrace yourself if it is.

If the error is from that warning, the code

np.array(predicted_labels, dtype=self._prediction_dtype).reshape(data.shape[0], -1)

looks like it's expecting the predicted_labels to have the same number of records of data (which is the request we sent to the model). This may mean if we sent N records to your model, but according to your analysis configuration, when we extract the predicted labels from your model output, the number of labels is different than N.

We use label and probability parameters under predictor in the Analysis Configuration to extract those predicted labels/scores. Your label config may be resulting in that "ragged nested sequences" the warning is talking about.

Please check to make sure your model + analysis config ensures there is only one predicted label per record and label is producing what you expect