{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": Infinity, "token_generation_latency_sync": 18.704550361633302, "token_generation_latency_async": 18.695793859660625, "token_generation_throughput_sync": 0.05346292643586857, "token_generation_throughput_async": 0.05348796673232856, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 195.85751037597657, "inference_latency_async": 26.712918281555176, "inference_throughput_sync": 0.005105752636599723, "inference_throughput_async": 0.03743507128124161, "inference_CO2_emissions": null, "inference_energy_consumption": null }