{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 501916.125, "token_generation_latency_sync": 167.35362243652344, "token_generation_latency_async": 167.56441667675972, "token_generation_throughput_sync": 0.0059753711060499815, "token_generation_throughput_async": 0.0059678541532421575, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 266.4074249267578, "inference_latency_async": 213.09516429901123, "inference_throughput_sync": 0.0037536491344973793, "inference_throughput_async": 0.004692739055292772, "inference_CO2_emissions": null, "inference_energy_consumption": null }