Apply the Metal kernel optimized attention

Apply the Metal kernel optimized attention

Add programs from async context - this simulates controller.run()

Calculate the Levenshtein edit distance between two code snippets Args: code1: First code snippet code2: Second code snippet

Cancel all tasks in the pool

Create hooks to replace Qwen3's attention with Metal kernel optimized version.

Create hooks to replace Qwen3's attention with Metal kernel optimized version.

Main stage evaluation with thorough testing on test dataset

Evaluate the program by running it multiple times and checking how close it gets to the known global minimum. Args: program_path

Evaluate the program by submitting it to OJ and fetching metrics based on how well it performs. Args: program_path: Path to the prog

Evaluate the web scraper program. Args: program_path: Path to the program to evaluate Returns: Dictionary with metrics

Main evaluation function called by OpenEvolve. IMPORTANT: OpenEvolve expects this exact function signature! It should return a dicti

Main evaluation function - for backwards compatibility Calls evaluate_stage2 for full evaluation Args: prompt_path: Path to the

Evaluate HotpotQA dataset.

Evaluate HoVer dataset.

Evaluate IFEval dataset.

Evaluate multiple programs in parallel Args: programs: List of (program_code, program_id) tuples Returns:

Stage 1 evaluation: Quick validation that the program runs without errors

First stage evaluation - quick validation check Enhanced with artifacts for debugging

First stage evaluation with basic functionality check of the evolved solve method

First stage evaluation with basic functionality check of the evolved solve method

First stage evaluation with basic functionality check of the evolved solve method

First stage evaluation with basic functionality check of the evolved solve method

First stage evaluation with basic functionality check of the evolved solve method

First stage evaluation with basic functionality check of the evolved solve method

First stage evaluation with basic functionality check of the evolved solve method

First stage evaluation with basic functionality check of the evolved solve method

Stage 1 evaluation: Quick evaluation with 10% of samples Args: prompt_path: Path to the prompt file Returns: Dictionary

First stage evaluation - quick validation check

First stage evaluation with fewer trials

Stage 2 evaluation: Full evaluation with all test signals

Second stage evaluation - full evaluation

Second stage evaluation with more thorough testing of the evolved solve method

Second stage evaluation with more thorough testing of the evolved solve method

Second stage evaluation with more thorough testing of the evolved solve method

Second stage evaluation with more thorough testing of the evolved solve method

Second stage evaluation with more thorough testing of the evolved solve method

Second stage evaluation with more thorough testing of the evolved solve method

Second stage evaluation with more thorough testing of the evolved solve method

Second stage evaluation with more thorough testing of the evolved solve method

Second stage evaluation - full evaluation

Second stage evaluation with more thorough testing

Get config for evolution tests

Create output directory for evolution tests

Simple evaluator for fibonacci function

Auto-wrap dict returns for backward compatibility

Run tasks with a concurrency limit Args: n: Maximum number of tasks to run concurrently *tasks: Tasks to run return_

Generate text from a prompt

Generate text from a prompt

Generate multiple texts in parallel

Mock async generate method that returns NOVEL

Generate text using a system message and conversational context

Get simple test messages for basic validation

Get detailed information about a task.

Get OpenAI client configured for local optillm

Construct an arrangement of n points on or inside a convex region in order to maximize the area of the smallest triangle formed by these poin

Construct an arrangement of n points on or inside a convex region in order to maximize the area of the smallest triangle formed by these poin

Construct an arrangement of n points on or inside a convex region in order to maximize the area of the smallest triangle formed by these poin

Constructs a packing of 11 disjoint unit regular hexagons inside a larger regular hexagon, maximizing 1/outer_hex_side_length. Returns

Constructs a packing of 12 disjoint unit regular hexagons inside a larger regular hexagon, maximizing 1/outer_hex_side_length. Returns

Test function that mimics the logic in openai.py

Check if the provided solution is valid. Args: problem: The original problem solution: The proposed

Check if the provided solution is valid. Args: problem: The original problem solution: The proposed