{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 8.611993789672852, "token_generation_latency_sync": 17.415401935577393, "token_generation_latency_async": 17.745299078524113, "token_generation_throughput_sync": 0.057420437593067, "token_generation_throughput_async": 0.05635295272144664, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 188.5877243041992, "inference_latency_async": 47.14810848236084, "inference_throughput_sync": 0.005302572071907298, "inference_throughput_async": 0.02120975861362757, "inference_CO2_emissions": null, "inference_energy_consumption": null }