Model import registration flow using Pipelines

[HrishikeshGeedMS]

Description

This is a sample notebook to use registered component available in prod registry , to create a pipeline and submit a job in workspace

Checklist

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

[ManojBableshwar]

@HrishikeshGeedMS @skasturi

looking at the latest component: https://ml.azure.com/registries/azureml-preview/components/download_model

model_source: type: string optional: False description: Supported model containers enum: - AzureBlob - GIT - Huggingface

can we make this optional and default to Huggingface?

[ManojBableshwar]

mlflow_flavor: type: string optional: False description: MLFlow flavor to translate model to. Flavor information is used to determine model loader module enum: - hftransformers

can this be automatically detected using the model's config.json? users may not easily get this. Additionally, we would like to use just transformers and not hftransformers

[ManojBableshwar]

Can we try if there is a way to avoid the task_type in model converter? This info can be fetched using HF hub api and saved in the model_info file.

[skasturi]

@HrishikeshGeedMS Is this PR still active?