{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 8.80232048034668, "token_generation_latency_sync": 53.172634506225585, "token_generation_latency_async": 53.34047880023718, "token_generation_throughput_sync": 0.018806666423175203, "token_generation_throughput_async": 0.0187474882583085, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 195.83047637939453, "inference_latency_async": 109.62216854095459, "inference_throughput_sync": 0.005106457475304497, "inference_throughput_async": 0.009122242456154317, "inference_CO2_emissions": null, "inference_energy_consumption": null }