{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 9.657100677490234, "token_generation_latency_sync": 276.73154296875, "token_generation_latency_async": 276.67152024805546, "token_generation_throughput_sync": 0.003613610466201626, "token_generation_throughput_async": 0.003614394423768047, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 265.28133087158204, "inference_latency_async": 197.52421379089355, "inference_throughput_sync": 0.0037695830185806864, "inference_throughput_async": 0.005062670448386835, "inference_CO2_emissions": null, "inference_energy_consumption": null }