{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 13647.91796875, "token_generation_latency_sync": 15.27142333984375, "token_generation_latency_async": 15.62624853104353, "token_generation_throughput_sync": 0.0654817810852614, "token_generation_throughput_async": 0.06399488642545093, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 14.187315273284913, "inference_latency_async": 14.705300331115723, "inference_throughput_sync": 0.07048549924614887, "inference_throughput_async": 0.06800269137543877, "inference_CO2_emissions": null, "inference_energy_consumption": null }