{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 171265408.0, "token_generation_latency_sync": 92.45938034057617, "token_generation_latency_async": 93.7770240008831, "token_generation_throughput_sync": 0.010815560263506828, "token_generation_throughput_async": 0.010663592821953734, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 94.79567337036133, "inference_latency_async": 95.85397243499756, "inference_throughput_sync": 0.010549004658609857, "inference_throughput_async": 0.010432535810429144, "inference_CO2_emissions": null, "inference_energy_consumption": null }