{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 11521722.0, "token_generation_latency_sync": 96.70892639160157, "token_generation_latency_async": 96.56033273786306, "token_generation_throughput_sync": 0.010340307118607847, "token_generation_throughput_async": 0.01035621949144218, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 170.99335632324218, "inference_latency_async": 137.30134963989258, "inference_throughput_sync": 0.00584818042935903, "inference_throughput_async": 0.007283249601134674, "inference_CO2_emissions": null, "inference_energy_consumption": null }