T5Transformer.pretrained("t5_grammar_error_corrector") - IllegalArgumentException: No Operation named [missing_encoder_input_ids] in the Graph

[ancil009]

Hi Team,

I am getting below error , while downloading "t5_grammar_error_corrector" model from T5Transformer pretrained model.
IllegalArgumentException: No Operation named [missing_encoder_input_ids] in the Graph

Same error in colab official notebook and databricks notebook.

Please suggest a solution for the same.................
Databricks
image

Colab
image

[maziyarpanahi]

Hi,

Thanks for reporting this. We need to re-import this model (the only T5 left behind) with new architecture we introduced in Spark NLP 4.x.

I will start the motion for this to happen soon this week.

PS: original T5 model: https://huggingface.co/prithivida/grammar_error_correcter_v1

[ancil009]

Thanks for quick response

[ancil009]

Hi @maziyarpanahi

I am changed sparknlp version to 3.4.4. then its working colab notebook. In my project we are using databricks. But in databricks notebook getting same error again

[ancil009]

In Cluster

[maziyarpanahi]

We will fix this specific model for Spark NLP 4.x soon this week. You shouldn't downgrade that far back in your project (different Apache Spark support, etc.).

Anything that is broken in Spark NLP 3.4.x won't be fixed, let's stay on 4.2.x and we will have the compatible version of this model soon.

[ancil009]

We will fix this specific model for Spark NLP 4.x soon this week. You shouldn't downgrade that far back in your project (different Apache Spark support, etc.).

Anything that is broken in Spark NLP 3.4.x won't be fixed, let's stay on 4.2.x and we will have the compatible version of this model soon.

Okay, Thanks a lot @maziyarpanahi