tfdv.validate_tensor_examples()?

[schmidt-jake]

I think it would be nice to have a top-level function to check for anomalies in serving data. It could be integrated into serving_input_receiver_fn. It doesn't make sense to have to write serving data to tfrecords just to make use of tfdv.generate_statistics_from_tfrecord.

[rmothukuru]

@JakeTheWise,
We currently, can check if Serving Data has Anomalies using the below code:

serving_stats = tfdv.generate_statistics_from_tfrecord(data_location=serving_data_path)
serving_anomalies = tfdv.validate_statistics(serving_stats, schema)

For more information, please refer this TFDV Page.

Please let me know if this is what you are looking for, or if your proposal is different. Thanks!

[schmidt-jake]

@rmothukuru Yes I was aware of generate_statistics_from_tfrecord (and I'll update the original post). Why do I need to write serving data to tfrecords just to check for anomalies?

[caveness]

Hi Jake -- Are you looking to validate your serving examples one at a time?

[schmidt-jake]

In the limit, yes. Serving examples could come in one at a time, or in batches. Thus the idea to have a tensor-based example validator, as the serving environment (tensorflow model server) / model export signature function will receive examples as (dictionaries of) tensor objects.