{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 1200.4630126953125, "token_generation_latency_sync": 167.0520217895508, "token_generation_latency_async": 167.15887673199177, "token_generation_throughput_sync": 0.005986159217275338, "token_generation_throughput_async": 0.005982332614039483, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 261.15655822753905, "inference_latency_async": 209.26249027252197, "inference_throughput_sync": 0.0038291207649042668, "inference_throughput_async": 0.004778687277866677, "inference_CO2_emissions": null, "inference_energy_consumption": null }