{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 7.370403289794922, "token_generation_latency_sync": 50.041364669799805, "token_generation_latency_async": 48.99688959121704, "token_generation_throughput_sync": 0.01998346780905247, "token_generation_throughput_async": 0.0204094588114274, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 192.5317642211914, "inference_latency_async": 104.32147979736328, "inference_throughput_sync": 0.005193948146920543, "inference_throughput_async": 0.009585753594968415, "inference_CO2_emissions": null, "inference_energy_consumption": null }