{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 8.037196159362793, "token_generation_latency_sync": 231.49828338623047, "token_generation_latency_async": 231.63429871201515, "token_generation_throughput_sync": 0.004319686458890952, "token_generation_throughput_async": 0.004317149945238782, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 326.71918029785155, "inference_latency_async": 212.95204162597656, "inference_throughput_sync": 0.0030607324586464624, "inference_throughput_async": 0.00469589299245308, "inference_CO2_emissions": null, "inference_energy_consumption": null }