{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 7.124024868011475, "token_generation_latency_sync": 16.533242607116698, "token_generation_latency_async": 16.764117032289505, "token_generation_throughput_sync": 0.06048420287315884, "token_generation_throughput_async": 0.05965121801964826, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 183.09539794921875, "inference_latency_async": 46.03679180145264, "inference_throughput_sync": 0.005461633723188109, "inference_throughput_async": 0.021721756900715358, "inference_CO2_emissions": null, "inference_energy_consumption": null }