{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 10.935739517211914, "token_generation_latency_sync": 71.9461898803711, "token_generation_latency_async": 71.44534755498171, "token_generation_throughput_sync": 0.013899276690854029, "token_generation_throughput_async": 0.013996712651309267, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 71.43147506713868, "inference_latency_async": 68.55134963989258, "inference_throughput_sync": 0.013999430909974865, "inference_throughput_async": 0.014587604842984197, "inference_CO2_emissions": null, "inference_energy_consumption": null }