{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 14.672518730163574, "token_generation_latency_sync": 18.520208168029786, "token_generation_latency_async": 18.467084877192974, "token_generation_throughput_sync": 0.053995073431530545, "token_generation_throughput_async": 0.05415039821661347, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 192.8057876586914, "inference_latency_async": 46.9390869140625, "inference_throughput_sync": 0.005186566296288883, "inference_throughput_async": 0.021304206488524803, "inference_CO2_emissions": null, "inference_energy_consumption": null }