Extract repeating question string into TEST_QUESTION variable and
create make_request() helper function to reduce code duplication.
Add proper error handling for error responses.
Add a standalone Python script that simulates a llama-server HTTP endpoint
for testing the eval script. The simulator:
- Implements /v1/chat/completions endpoint with OpenAI-compatible format
- Loads AIME dataset from HuggingFace with local caching
- Uses Levenshtein distance for intelligent question matching
- Supports configurable success rate for correct/wrong answer generation
- Provides debug logging for troubleshooting
Also includes test scripts and documentation for testing and understanding
the simulator functionality.
Georgi Gerganov