{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 122000.8671875, "token_generation_latency_sync": 174.83052520751954, "token_generation_latency_async": 174.52176827937365, "token_generation_throughput_sync": 0.005719824949407574, "token_generation_throughput_async": 0.005729944234802873, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 270.43553466796874, "inference_latency_async": 232.04243183135986, "inference_throughput_sync": 0.0036977389130010774, "inference_throughput_async": 0.00430955662767215, "inference_CO2_emissions": null, "inference_energy_consumption": null }