{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 9.663578033447266, "token_generation_latency_sync": 62.545572662353514, "token_generation_latency_async": 62.48115599155426, "token_generation_throughput_sync": 0.015988341899088643, "token_generation_throughput_async": 0.016004825521076668, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 196.81443786621094, "inference_latency_async": 110.87660789489746, "inference_throughput_sync": 0.005080928060164837, "inference_throughput_async": 0.009019034934293116, "inference_CO2_emissions": null, "inference_energy_consumption": null }