{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 14.67513656616211, "token_generation_latency_sync": 36.57426109313965, "token_generation_latency_async": 36.32111717015505, "token_generation_throughput_sync": 0.027341632342302417, "token_generation_throughput_async": 0.027532192782376664, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 51.19948806762695, "inference_latency_async": 46.827030181884766, "inference_throughput_sync": 0.019531445288654994, "inference_throughput_async": 0.02135518729494091, "inference_CO2_emissions": null, "inference_energy_consumption": null }