{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 9.242453575134277, "token_generation_latency_sync": 166.1800277709961, "token_generation_latency_async": 166.38644598424435, "token_generation_throughput_sync": 0.006017570302600064, "token_generation_throughput_async": 0.00601010493423661, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 266.1434356689453, "inference_latency_async": 197.44746685028076, "inference_throughput_sync": 0.0037573724014140094, "inference_throughput_async": 0.005064638285576354, "inference_CO2_emissions": null, "inference_energy_consumption": null }