{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 18.01762580871582, "token_generation_latency_sync": 51.89610748291015, "token_generation_latency_async": 51.79595183581114, "token_generation_throughput_sync": 0.0192692679374712, "token_generation_throughput_async": 0.01930652810802506, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 197.18768768310548, "inference_latency_async": 111.5454912185669, "inference_throughput_sync": 0.00507131054555024, "inference_throughput_async": 0.008964952227791603, "inference_CO2_emissions": null, "inference_energy_consumption": null }