{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 353653.375, "token_generation_latency_sync": 166.52535552978514, "token_generation_latency_async": 166.5806146338582, "token_generation_throughput_sync": 0.006005091517856795, "token_generation_throughput_async": 0.0060030994734770644, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 262.70494689941404, "inference_latency_async": 209.80782508850098, "inference_throughput_sync": 0.0038065518438177167, "inference_throughput_async": 0.004766266461120698, "inference_CO2_emissions": null, "inference_energy_consumption": null }