{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 231163.296875, "token_generation_latency_sync": 132.18108673095702, "token_generation_latency_async": 132.18097016215324, "token_generation_throughput_sync": 0.0075653788657027154, "token_generation_throughput_async": 0.007565385537519116, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 202.5590789794922, "inference_latency_async": 170.84074020385742, "inference_throughput_sync": 0.004936831294050481, "inference_throughput_async": 0.005853404748813076, "inference_CO2_emissions": null, "inference_energy_consumption": null }