{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 7.146307468414307, "token_generation_latency_sync": 16.625248146057128, "token_generation_latency_async": 16.69239643961191, "token_generation_throughput_sync": 0.06014947814400964, "token_generation_throughput_async": 0.05990751559356383, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 188.58076019287108, "inference_latency_async": 46.52762413024902, "inference_throughput_sync": 0.005302767890940992, "inference_throughput_async": 0.021492608287941133, "inference_CO2_emissions": null, "inference_energy_consumption": null }