{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 266904.21875, "token_generation_latency_sync": 287.5701171875, "token_generation_latency_async": 287.3256243765354, "token_generation_throughput_sync": 0.0034774127777260845, "token_generation_throughput_async": 0.0034803717982685623, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 277.70613861083984, "inference_latency_async": 267.5260066986084, "inference_throughput_sync": 0.00360092868311182, "inference_throughput_async": 0.0037379543482162765, "inference_CO2_emissions": null, "inference_energy_consumption": null }