Added adopter to test llms for real-word-dataset : ManyType4Py

[rashidabhar]

This merge request includes following

1. Updating current llm adopter to allow masked based prompting
2. Real-world-llm adopter to run data preprocessing, fine_tuning and model runners

[Copilot]

Pull Request Overview

This PR adds support for a new LLM adopter for processing the ManyTypes4Py real-world dataset. Key changes include:

• A new dataset preprocessing module that performs type hint stripping, file categorization, and JSON updates.
• A new code annotator module that masks type annotations using LibCST.
• Updates to runners, prompt templates, and result translation logic for LLM inference, alongside documentation and configuration improvements.

Reviewed Changes

Copilot reviewed 36 out of 41 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
src/target_tools/real-world-llms/src/dataset-preprocessing/prepare_dataset.py	Introduces dataset preprocessing and JSON update logic with type-cleaning and split categorization.
src/target_tools/real-world-llms/src/code_annotator.py	Implements a CST transformer to mask type annotations in function definitions and assignments.
src/target_tools/real-world-llms/models_info.md	Adds a comprehensive list of supported LLM models.
src/target_tools/real-world-llms/README.md	Updates instructions for dataset preprocessing and model inference.
src/target_tools/llms/src/utils.py	Adds handling for masked code-based prompt generation.
src/target_tools/llms/src/runner.py	Incorporates a new JSON creation function for code files and integrates result translation.
src/target_tools/llms/src/result_translator.py	Provides functionality to translate raw code outputs into JSON-formatted type annotations.
src/target_tools/llms/src/prompts.py	Introduces new prompt templates for masked code based tasks.
src/target_tools/hityperdl/*	Includes updates for type normalization and runner adjustments for HiTyper integration.
src/runner_class.py	Updates Docker container creation with an Nvidia runtime specification.
src/result_analyzer/*	Adds new analysis utilities for large-scale result matching and evaluation.

Files not reviewed (5)

• src/target_tools/hityper/Dockerfile: Language not supported
• src/target_tools/hityperdl/Dockerfile.cuda: Language not supported
• src/target_tools/hityperdl/requirements.txt: Language not supported
• src/target_tools/real-world-llms/Dockerfile: Language not supported
• src/target_tools/real-world-llms/requirements.txt: Language not supported

[Copilot]

The global variable 'libcst_fail_count' is used without being initialized in the module; consider initializing it at the module level to ensure consistent behavior.

[Copilot]

The flag '--bechmark_path' appears to be misspelled; consider renaming it to '--benchmark_path' to avoid confusion.

Suggested change

	--bechmark_path /mnt/hf_cache/rashida_manytype4py/many-types-4-py-dataset/rw-benchmark \
	--benchmark_path /mnt/hf_cache/rashida_manytype4py/many-types-4-py-dataset/rw-benchmark \

[Copilot]

The key 'dictonary' appears to be a typo; consider correcting it to 'dictionary' for clarity and consistency.

Suggested change

	"dictonary": "dict",
	"dictionary": "dict",