{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 251299.453125, "token_generation_latency_sync": 91.27292938232422, "token_generation_latency_async": 91.43062066286802, "token_generation_throughput_sync": 0.01095615103807174, "token_generation_throughput_async": 0.010937254857836943, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 86.61442642211914, "inference_latency_async": 88.07699680328369, "inference_throughput_sync": 0.01154542079545106, "inference_throughput_async": 0.011353702286573853, "inference_CO2_emissions": null, "inference_energy_consumption": null }