Add gemini support for batch mode

[aashay-sarvam]

Support gemini with batch -
https://cloud.google.com/vertex-ai/generative-ai/docs/model-reference/batch-prediction-api#generative-ai-batch-text-

[RyanMarten]

Related:

• [Doc] https://cloud.google.com/vertex-ai/generative-ai/docs/multimodal/batch-prediction-gemini
• [Example Code] https://github.com/GoogleCloudPlatform/generative-ai/blob/main/gemini/batch-prediction/intro_batch_prediction.ipynb
• [Example Code] https://github.com/GoogleCloudPlatform/generative-ai/blob/main/gemini/batch-prediction/intro_batch_prediction_using_bigquery_input.ipynb

@aashay-sarvam Thanks for the suggestion :) We welcome contributions - if you are interested in implementing this in a PR, I can give some pointers in the code where to start (should be easy with the abstractions we have for the batch request processor).

Based on the examples above, it looks like Google has an existing python API for this that we can leverage from vertexai.batch_prediction import BatchPredictionJob

submit:

job = BatchPredictionJob.submit(source_model=MODEL_ID, input_dataset=INPUT_DATA,output_uri_prefix=BUCKET_URI`)

poll:

job.has_ended

job.has_succeeded

delete:

job.delete()

upload (and deletion) of batch file and download of batch results with from google.cloud import storage

[RyanMarten]

Code reference for BatchPredictionJob

https://github.com/googleapis/python-aiplatform/blob/de85a0c0d5ffaf0713279923c1e3619cba84e087/vertexai/batch_prediction/_batch_prediction.py#L41

https://github.com/googleapis/python-aiplatform/blob/de85a0c0d5ffaf0713279923c1e3619cba84e087/google/cloud/aiplatform_v1/types/batch_prediction_job.py#L47