{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 6.429513454437256, "token_generation_latency_sync": 468.8035583496094, "token_generation_latency_async": 468.5971027240157, "token_generation_throughput_sync": 0.0021330896111804933, "token_generation_throughput_async": 0.0021340294128727438, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 450.4276947021484, "inference_latency_async": 360.5192184448242, "inference_throughput_sync": 0.002220112155095756, "inference_throughput_async": 0.002773777232497372, "inference_CO2_emissions": null, "inference_energy_consumption": null }