{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 13603.970703125, "token_generation_latency_sync": 18.421577072143556, "token_generation_latency_async": 18.873421289026737, "token_generation_throughput_sync": 0.05428416883547739, "token_generation_throughput_async": 0.05298456409603984, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 17.43585271835327, "inference_latency_async": 17.00263023376465, "inference_throughput_sync": 0.05735308826894272, "inference_throughput_async": 0.05881442966477925, "inference_CO2_emissions": null, "inference_energy_consumption": null }