Create zero_delay_qkv_benchmark.py

[itsmenick212]

Implemented Zero-Delay QKV Compression with benchmarking for HuggingFace Transformers to optimize inference runtime.

• Added CombinedQKVProjection class to replace separate query, key, and value projections with a single linear layer.
• Modified BERT's attention mechanism by replacing query, key, and value layers with the combined QKV projection.
• Introduced a replace_attn_qkv_with_combined function to automate the replacement process across all layers.
• Added a benchmarking utility to compare inference runtime performance before and after QKV compression.
• Provided a complete example script for testing and benchmarking with dummy data.

This change aims to demonstrate runtime optimizations for inference-heavy use cases in Transformer models.

What does this PR do?

Implemented Zero-Delay QKV Compression with benchmarking for HuggingFace Transformers to optimize inference runtime.

• Added CombinedQKVProjection class to replace separate query, key, and value projections with a single linear layer.
• Modified BERT's attention mechanism by replacing query, key, and value layers with the combined QKV projection.
• Introduced a replace_attn_qkv_with_combined function to automate the replacement process across all layers.
• Added a benchmarking utility to compare inference runtime performance before and after QKV compression.
• Provided a complete example script for testing and benchmarking with dummy data.

This change aims to demonstrate runtime optimizations for inference-heavy use cases in Transformer models.

Fixes # (issue)

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[itsmenick212]

Implemented Zero-Delay QKV Compression with benchmarking for HuggingFace Transformers to optimize inference runtime.

• Added CombinedQKVProjection class to replace separate query, key, and value projections with a single linear layer.
• Modified BERT's attention mechanism by replacing query, key, and value layers with the combined QKV projection.
• Introduced a replace_attn_qkv_with_combined function to automate the replacement process across all layers.
• Added a benchmarking utility to compare inference runtime performance before and after QKV compression.
• Provided a complete example script for testing and benchmarking with dummy data.

This change aims to demonstrate runtime optimizations for inference-heavy use cases in Transformer models.