{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 23302.48828125, "token_generation_latency_sync": 297.8766571044922, "token_generation_latency_async": 298.0090944096446, "token_generation_throughput_sync": 0.0033570942071140873, "token_generation_throughput_async": 0.0033556022912018773, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 459.06605834960936, "inference_latency_async": 411.5968704223633, "inference_throughput_sync": 0.0021783357358091445, "inference_throughput_async": 0.0024295617189067603, "inference_CO2_emissions": null, "inference_energy_consumption": null }