{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 102042.3359375, "token_generation_latency_sync": 268.81107940673826, "token_generation_latency_async": 269.08264122903347, "token_generation_throughput_sync": 0.0037200847606690315, "token_generation_throughput_async": 0.0037163304010712304, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 260.66544647216796, "inference_latency_async": 225.5446434020996, "inference_throughput_sync": 0.003836335093638017, "inference_throughput_async": 0.00443371203552463, "inference_CO2_emissions": null, "inference_energy_consumption": null }