results.json

{
    "current_gpu_type": "NVIDIA A100-PCIE-40GB",
    "current_gpu_total_memory": 40339.3125,
    "perplexity": Infinity,
    "token_generation_latency_sync": 9.605268955230713,
    "token_generation_latency_async": 9.503808431327343,
    "token_generation_throughput_sync": 0.10410952620493077,
    "token_generation_throughput_async": 0.10522097611981597,
    "token_generation_CO2_emissions": null,
    "token_generation_energy_consumption": null,
    "inference_latency_sync": 54.57582015991211,
    "inference_latency_async": 17.85736083984375,
    "inference_throughput_sync": 0.018323132791589924,
    "inference_throughput_async": 0.05599931641459455,
    "inference_CO2_emissions": null,
    "inference_energy_consumption": null
}