{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 77205.703125, "token_generation_latency_sync": 175.1979782104492, "token_generation_latency_async": 175.4031714051962, "token_generation_throughput_sync": 0.005707828424816592, "token_generation_throughput_async": 0.005701151193497609, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 267.9752700805664, "inference_latency_async": 232.43465423583984, "inference_throughput_sync": 0.0037316876281134124, "inference_throughput_async": 0.004302284456195374, "inference_CO2_emissions": null, "inference_energy_consumption": null }