{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 8.28774356842041, "token_generation_latency_sync": 56.85212440490723, "token_generation_latency_async": 56.51623625308275, "token_generation_throughput_sync": 0.017589492221572717, "token_generation_throughput_async": 0.01769403035831944, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 192.63211822509766, "inference_latency_async": 107.73584842681885, "inference_throughput_sync": 0.00519124229756672, "inference_throughput_async": 0.009281961525362328, "inference_CO2_emissions": null, "inference_energy_consumption": null }