{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 215362.296875, "token_generation_latency_sync": 282.7625244140625, "token_generation_latency_async": 281.9853914901614, "token_generation_throughput_sync": 0.003536536540944347, "token_generation_throughput_async": 0.0035462829996811746, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 272.0771057128906, "inference_latency_async": 234.2961072921753, "inference_throughput_sync": 0.003675428689157147, "inference_throughput_async": 0.004268103348183099, "inference_CO2_emissions": null, "inference_energy_consumption": null }