{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 6.575528621673584, "token_generation_latency_sync": 154.8782928466797, "token_generation_latency_async": 155.20499553531408, "token_generation_throughput_sync": 0.006456682738554851, "token_generation_throughput_async": 0.0064430915805958585, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 245.54997711181642, "inference_latency_async": 183.94174575805664, "inference_throughput_sync": 0.004072490707440093, "inference_throughput_async": 0.005436503801129114, "inference_CO2_emissions": null, "inference_energy_consumption": null }