{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 472478.8125, "token_generation_latency_sync": 29.096915626525877, "token_generation_latency_async": 28.846427984535694, "token_generation_throughput_sync": 0.034367903898664814, "token_generation_throughput_async": 0.034666337216382244, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 27.447398376464843, "inference_latency_async": 27.158403396606445, "inference_throughput_sync": 0.03643332552995129, "inference_throughput_async": 0.036821015779040754, "inference_CO2_emissions": null, "inference_energy_consumption": null }