{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 1567938.375, "token_generation_latency_sync": 100.42423782348632, "token_generation_latency_async": 99.31453671306372, "token_generation_throughput_sync": 0.00995775543507415, "token_generation_throughput_async": 0.010069019431557809, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 92.28462142944336, "inference_latency_async": 95.04306316375732, "inference_throughput_sync": 0.010836041634136786, "inference_throughput_async": 0.01052154640972608, "inference_CO2_emissions": null, "inference_energy_consumption": null }