{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 9.87337875366211, "token_generation_latency_sync": 166.51768493652344, "token_generation_latency_async": 166.50729943066835, "token_generation_throughput_sync": 0.006005368140814593, "token_generation_throughput_async": 0.00600574271169648, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 267.5055618286133, "inference_latency_async": 198.30944538116455, "inference_throughput_sync": 0.003738240031961222, "inference_throughput_async": 0.005042624157805144, "inference_CO2_emissions": null, "inference_energy_consumption": null }