src/utils.py

# Example result structure for vLLM benchmarks
RESULT_SCHEMA = {
    "run_id": {
        "model": "model_name",
        "timestamp": "YYYY-MM-DD HH:MM:SS",
        "config": {
            "input_length": 128,
            "output_length": 128,
            "concurrent_requests": 16
        },
        "performance": {
            "median_ttft_ms": 0.0,      # Time To First Token
            "median_tpot_ms": 0.0,      # Time Per Output Token
            "median_itl_ms": 0.0,       # Inter-Token Latency
            "median_e2el_ms": 0.0,      # End-to-End Latency
            "total_token_throughput": 0.0
        },
        "accuracy": {
            "wikitext_perplexity": 0.0
        }
    }
}

# Mock data for testing
MOCK_RESULTS = {
    "run_2024_12_01_baseline": {
        "model": "amd/Mixtral-8x7B-Instruct-v0.1-FP8-KV",
        "timestamp": "2024-12-01 10:30:00",
        "config": {
            "input_length": 128,
            "output_length": 128,
            "concurrent_requests": 16
        },
        "performance": {
            "median_ttft_ms": 45.23,
            "median_tpot_ms": 12.56,
            "median_itl_ms": 11.89,
            "median_e2el_ms": 1589.45,
            "total_token_throughput": 2048.67
        },
        "accuracy": {
            "wikitext_perplexity": 7.89
        }
    },
    "run_2024_12_02_optimized": {
        "model": "amd/Mixtral-8x7B-Instruct-v0.1-FP8-KV",
        "timestamp": "2024-12-02 14:15:00",
        "config": {
            "input_length": 128,
            "output_length": 128,
            "concurrent_requests": 16
        },
        "performance": {
            "median_ttft_ms": 42.11,
            "median_tpot_ms": 11.23,
            "median_itl_ms": 10.95,
            "median_e2el_ms": 1456.78,
            "total_token_throughput": 2234.89
        },
        "accuracy": {
            "wikitext_perplexity": 7.91
        }
    }
}