{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 6.638092517852783, "token_generation_latency_sync": 57.643031692504884, "token_generation_latency_async": 57.63615034520626, "token_generation_throughput_sync": 0.0173481506894792, "token_generation_throughput_async": 0.01735022193554904, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 184.3141616821289, "inference_latency_async": 105.04143238067627, "inference_throughput_sync": 0.005425519075005293, "inference_throughput_async": 0.009520052967061052, "inference_CO2_emissions": null, "inference_energy_consumption": null }