{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 9.425127983093262, "token_generation_latency_sync": 270.3133880615234, "token_generation_latency_async": 270.79987842589617, "token_generation_throughput_sync": 0.00369940981159394, "token_generation_throughput_async": 0.0036927638439603213, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 259.8624267578125, "inference_latency_async": 192.05210208892822, "inference_throughput_sync": 0.003848190030688752, "inference_throughput_async": 0.005206920357148488, "inference_CO2_emissions": null, "inference_energy_consumption": null }