{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 34.23958969116211, "token_generation_latency_sync": 15.396154880523682, "token_generation_latency_async": 15.528209507465363, "token_generation_throughput_sync": 0.06495128217143437, "token_generation_throughput_async": 0.06439892503506207, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 15.73519344329834, "inference_latency_async": 14.961576461791992, "inference_throughput_sync": 0.06355180847337487, "inference_throughput_async": 0.06683787651346382, "inference_CO2_emissions": null, "inference_energy_consumption": null }