updating model import component limitations

[ShipraJain01]

Description

updating model import component limitations. Changes in 'import_model_into_registry.ipynb' and adding new 'requirements.txt' file

Checklist

• I have read the contribution guidelines
• Pull request includes test coverage for the included changes.

[ManojBableshwar]

we can do some wordsmithing but can get that post //build. the changes look good to me.

About additional packages, i have noticed that these are the additional packages needed for top 1000 models on HF

• You need to install sacremoses to use XLMTokenizer and BioGptTokenizer
• You need to install fugashi to use MecabTokenizer
• google-tapas-small-finetuned-wtq requires the torch-scatter library

[ShipraJain01]

we can do some wordsmithing but can get that post //build. the changes look good to me.

About additional packages, i have noticed that these are the additional packages needed for top 1000 models on HF

• You need to install sacremoses to use XLMTokenizer and BioGptTokenizer
• You need to install fugashi to use MecabTokenizer
• google-tapas-small-finetuned-wtq requires the torch-scatter library

hi @ManojBableshwar
Could you please share pinned version of these dependencies. Also these dependencies are coming from azureml-evaluate-mlflow and Model Import package. If we want to add additional dependencies, we will have to rebuild environments and do a fresh py-pi release. Do we need that now?

[ManojBableshwar]

the above packages impact a tiny percentage of models. pls log a task to revisit this post build