create endpoint with InferenceAmiVersion

[CoderHam]

This PR introduces several changes to the client.py file, primarily focusing on the create_endpoint function. The modifications aim to enhance the deployment process for both fine-tuned and pre-trained models, ensuring compatibility with various Python versions and providing a more robust error-handling mechanism.

Summary of Changes:

• Import Updates: The import statements have been streamlined, removing redundant imports from the typing module.
• create_endpoint Function:
  • A new variable useBoto is introduced to determine the deployment method based on the presence of s3_models_dir.
  • The function now includes a try-except block to handle potential errors when deleting the model.
  • The logic for setting the role variable has been improved, considering the useBoto flag.
  • The model deployment process has been updated to handle validation parameter issues in Python 3.6.
  • For pre-trained models, the PR adds a new deployment method using boto to set the InferenceAmiVersion.
• S3 URL Parsing: The PR replaces the parse_s3_url function with lazy_sagemaker().s3.parse_s3_url for parsing S3 URLs, ensuring consistency and compatibility.

Detailed Changes:

• Import Updates:
  • Removed: TokenLikelihood from the generation module.
  • Removed: Tuple from the typing module.
• create_endpoint Function:
  • Added: useBoto variable to determine the deployment method.
  • Added: try-except block to handle potential errors when deleting the model.
  • Modified: The logic for setting the role variable now considers the useBoto flag.
  • Modified: The model deployment process now handles validation parameter issues in Python 3.6.
  • Added: A new deployment method for pre-trained models using boto to set the InferenceAmiVersion.
• S3 URL Parsing:
  • Replaced: parse_s3_url with lazy_sagemaker().s3.parse_s3_url for parsing S3 URLs.