{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 842026.75, "token_generation_latency_sync": 121.19252243041993, "token_generation_latency_async": 121.19309064000845, "token_generation_throughput_sync": 0.008251334157799451, "token_generation_throughput_async": 0.008251295471706359, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 170.88665924072265, "inference_latency_async": 167.72325038909912, "inference_throughput_sync": 0.005851831877591635, "inference_throughput_async": 0.005962202602680977, "inference_CO2_emissions": null, "inference_energy_consumption": null }