{ "current_gpu_type": "NVIDIA A100-PCIE-40GB", "current_gpu_total_memory": 40339.3125, "perplexity": 216.64439392089844, "token_generation_latency_sync": 292.3894104003906, "token_generation_latency_async": 292.4048237502575, "token_generation_throughput_sync": 0.003420096502915839, "token_generation_throughput_async": 0.003419916221539828, "token_generation_CO2_emissions": null, "token_generation_energy_consumption": null, "inference_latency_sync": 457.3069366455078, "inference_latency_async": 387.743878364563, "inference_throughput_sync": 0.0021867151356489774, "inference_throughput_async": 0.0025790220189105964, "inference_CO2_emissions": null, "inference_energy_consumption": null }