| [ |
| { |
| "axis": "00_dataset_preflight_alpaca", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/00_dataset_preflight_alpaca/llm_gpt2_gpu0_ds_alpaca_fixed_maxtok8_nocap", |
| "run_name": "llm_gpt2_gpu0_ds_alpaca_fixed_maxtok8_nocap", |
| "model": "gpt2", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 8, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.019975, |
| "inference_energy_j": 0.288022, |
| "inference_power_avg_w": 14.419, |
| "inference_power_peak_w": 14.419, |
| "inference_prefill_duration_s": 0.21, |
| "inference_prefill_energy_j": 3.027984, |
| "inference_prefill_power_avg_w": 14.419, |
| "inference_prefill_power_peak_w": 14.419, |
| "inference_decode_duration_s": 0.030035, |
| "inference_decode_energy_j": 0.433074, |
| "inference_decode_power_avg_w": 14.419, |
| "inference_decode_power_peak_w": 14.419, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 0.0, |
| "inference_idle_energy_j": 0.0, |
| "inference_idle_power_avg_w": 0.0, |
| "inference_idle_power_peak_w": 0.0, |
| "idle_pre_duration_s": 30.000115, |
| "idle_pre_energy_j": 1603.328479, |
| "idle_pre_power_avg_w": 53.444011, |
| "idle_pre_power_peak_w": 79.267, |
| "idle_mid_duration_s": 32.519979, |
| "idle_mid_energy_j": 318.662017, |
| "idle_mid_power_avg_w": 9.798952, |
| "idle_mid_power_peak_w": 14.791, |
| "idle_post_duration_s": 29.989963, |
| "idle_post_energy_j": 1580.137367, |
| "idle_post_power_avg_w": 52.674196, |
| "idle_post_power_peak_w": 78.677, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 1.0, |
| "total_input_tokens": 7.0, |
| "total_output_tokens": 8.0, |
| "prefill_time_total_ms": 205.0161361694336, |
| "decode_time_total_ms": 32.6995849609375, |
| "prefill_time_avg_ms": 205.0161361694336, |
| "decode_time_avg_ms": 32.6995849609375, |
| "decode_tokens_per_s": 244.65142323845077, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 2.956127667427063, |
| "est_decode_energy_j": 0.47149531555175783, |
| "est_total_infer_energy_j": 3.4276229829788205, |
| "est_j_per_output_token": 0.42845287287235256 |
| }, |
| { |
| "axis": "00_dataset_preflight_longbench", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/00_dataset_preflight_longbench/llm_gpt2_gpu0_ds_longbench_fixed_maxtok8_nocap", |
| "run_name": "llm_gpt2_gpu0_ds_longbench_fixed_maxtok8_nocap", |
| "model": "gpt2", |
| "prompt_set": "longbench", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 8, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.042038, |
| "inference_energy_j": 0.352909, |
| "inference_power_avg_w": 8.395, |
| "inference_power_peak_w": 8.395, |
| "inference_prefill_duration_s": 0.169986, |
| "inference_prefill_energy_j": 1.427029, |
| "inference_prefill_power_avg_w": 8.395, |
| "inference_prefill_power_peak_w": 8.395, |
| "inference_decode_duration_s": 0.080003, |
| "inference_decode_energy_j": 0.671622, |
| "inference_decode_power_avg_w": 8.395, |
| "inference_decode_power_peak_w": 8.395, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 0.0, |
| "inference_idle_energy_j": 0.0, |
| "inference_idle_power_avg_w": 0.0, |
| "inference_idle_power_peak_w": 0.0, |
| "idle_pre_duration_s": 30.01007, |
| "idle_pre_energy_j": 1644.34956, |
| "idle_pre_power_avg_w": 54.793168, |
| "idle_pre_power_peak_w": 79.372, |
| "idle_mid_duration_s": 34.127366, |
| "idle_mid_energy_j": 333.663614, |
| "idle_mid_power_avg_w": 9.777326, |
| "idle_mid_power_peak_w": 14.891, |
| "idle_post_duration_s": 29.989997, |
| "idle_post_energy_j": 1565.099801, |
| "idle_post_power_avg_w": 52.172607, |
| "idle_post_power_peak_w": 80.133, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 1.0, |
| "total_input_tokens": 512.0, |
| "total_output_tokens": 8.0, |
| "prefill_time_total_ms": 171.48399353027344, |
| "decode_time_total_ms": 74.22256469726562, |
| "prefill_time_avg_ms": 171.48399353027344, |
| "decode_time_avg_ms": 74.22256469726562, |
| "decode_tokens_per_s": 107.78393380274451, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 1.4396081256866455, |
| "est_decode_energy_j": 0.6230984306335449, |
| "est_total_infer_energy_j": 2.0627065563201903, |
| "est_j_per_output_token": 0.2578383195400238 |
| }, |
| { |
| "axis": "00_dataset_preflight_mmlu-pro", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/00_dataset_preflight_mmlu-pro/llm_gpt2_gpu0_ds_mmlu-pro_fixed_maxtok8_nocap", |
| "run_name": "llm_gpt2_gpu0_ds_mmlu-pro_fixed_maxtok8_nocap", |
| "model": "gpt2", |
| "prompt_set": "mmlu-pro", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 8, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.01989, |
| "inference_energy_j": 0.300305, |
| "inference_power_avg_w": 15.098, |
| "inference_power_peak_w": 15.098, |
| "inference_prefill_duration_s": 0.17001, |
| "inference_prefill_energy_j": 2.566806, |
| "inference_prefill_power_avg_w": 15.098, |
| "inference_prefill_power_peak_w": 15.098, |
| "inference_decode_duration_s": 0.059992, |
| "inference_decode_energy_j": 0.905765, |
| "inference_decode_power_avg_w": 15.098, |
| "inference_decode_power_peak_w": 15.098, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 0.0, |
| "inference_idle_energy_j": 0.0, |
| "inference_idle_power_avg_w": 0.0, |
| "inference_idle_power_peak_w": 0.0, |
| "idle_pre_duration_s": 30.000103, |
| "idle_pre_energy_j": 1647.366947, |
| "idle_pre_power_avg_w": 54.912037, |
| "idle_pre_power_peak_w": 78.797, |
| "idle_mid_duration_s": 32.530073, |
| "idle_mid_energy_j": 327.709257, |
| "idle_mid_power_avg_w": 10.074017, |
| "idle_mid_power_peak_w": 15.098, |
| "idle_post_duration_s": 29.990102, |
| "idle_post_energy_j": 1613.305145, |
| "idle_post_power_avg_w": 53.779683, |
| "idle_post_power_peak_w": 80.439, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 1.0, |
| "total_input_tokens": 187.0, |
| "total_output_tokens": 8.0, |
| "prefill_time_total_ms": 161.78297996520996, |
| "decode_time_total_ms": 66.1475658416748, |
| "prefill_time_avg_ms": 161.78297996520996, |
| "decode_time_avg_ms": 66.1475658416748, |
| "decode_tokens_per_s": 120.94171415389827, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 2.44259943151474, |
| "est_decode_energy_j": 0.9986959490776063, |
| "est_total_infer_energy_j": 3.4412953805923463, |
| "est_j_per_output_token": 0.4301619225740433 |
| }, |
| { |
| "axis": "00_smoke_test", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/00_smoke_test/llm_gpt2_gpu0_ds_alpaca_fixed_maxtok16_nocap", |
| "run_name": "llm_gpt2_gpu0_ds_alpaca_fixed_maxtok16_nocap", |
| "model": "gpt2", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 16, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.020489, |
| "inference_energy_j": 0.301808, |
| "inference_power_avg_w": 14.73, |
| "inference_power_peak_w": 14.73, |
| "inference_prefill_duration_s": 0.189995, |
| "inference_prefill_energy_j": 3.088379, |
| "inference_prefill_power_avg_w": 16.255, |
| "inference_prefill_power_peak_w": 20.525, |
| "inference_decode_duration_s": 0.059996, |
| "inference_decode_energy_j": 1.231422, |
| "inference_decode_power_avg_w": 20.525, |
| "inference_decode_power_peak_w": 20.525, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 0.0, |
| "inference_idle_energy_j": 0.0, |
| "inference_idle_power_avg_w": 0.0, |
| "inference_idle_power_peak_w": 0.0, |
| "idle_pre_duration_s": 30.01007, |
| "idle_pre_energy_j": 1567.207726, |
| "idle_pre_power_avg_w": 52.222543, |
| "idle_pre_power_peak_w": 79.077, |
| "idle_mid_duration_s": 32.542551, |
| "idle_mid_energy_j": 315.063969, |
| "idle_mid_power_avg_w": 9.681512, |
| "idle_mid_power_peak_w": 15.206, |
| "idle_post_duration_s": 29.990151, |
| "idle_post_energy_j": 1579.788627, |
| "idle_post_power_avg_w": 52.662576, |
| "idle_post_power_peak_w": 81.762, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 1.0, |
| "total_input_tokens": 7.0, |
| "total_output_tokens": 16.0, |
| "prefill_time_total_ms": 194.95677947998047, |
| "decode_time_total_ms": 53.414344787597656, |
| "prefill_time_avg_ms": 194.95677947998047, |
| "decode_time_avg_ms": 53.414344787597656, |
| "decode_tokens_per_s": 299.54500169615596, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 3.1690224504470823, |
| "est_decode_energy_j": 1.0963294267654418, |
| "est_total_infer_energy_j": 4.265351877212524, |
| "est_j_per_output_token": 0.26658449232578274 |
| }, |
| { |
| "axis": "01_model_scaling_initial", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/01_model_scaling_initial/llm_gpt2_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "run_name": "llm_gpt2_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "model": "gpt2", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.0, |
| "inference_energy_j": 0.0, |
| "inference_power_avg_w": 0.0, |
| "inference_power_peak_w": 0.0, |
| "inference_prefill_duration_s": 0.390446, |
| "inference_prefill_energy_j": 15.008107, |
| "inference_prefill_power_avg_w": 38.615103, |
| "inference_prefill_power_peak_w": 78.664, |
| "inference_decode_duration_s": 10.159686, |
| "inference_decode_energy_j": 785.23359, |
| "inference_decode_power_avg_w": 77.289164, |
| "inference_decode_power_peak_w": 92.152, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.051112, |
| "inference_idle_energy_j": 11951.016683, |
| "inference_idle_power_avg_w": 82.391827, |
| "inference_idle_power_peak_w": 100.889, |
| "idle_pre_duration_s": 30.010087, |
| "idle_pre_energy_j": 1607.821557, |
| "idle_pre_power_avg_w": 53.575974, |
| "idle_pre_power_peak_w": 78.753, |
| "idle_mid_duration_s": 32.509978, |
| "idle_mid_energy_j": 328.723316, |
| "idle_mid_power_avg_w": 10.11145, |
| "idle_mid_power_peak_w": 15.123, |
| "idle_post_duration_s": 29.989968, |
| "idle_post_energy_j": 1700.086444, |
| "idle_post_power_avg_w": 56.672675, |
| "idle_post_power_peak_w": 98.44, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 382.0, |
| "total_output_tokens": 3840.0, |
| "prefill_time_total_ms": 351.2871265411377, |
| "decode_time_total_ms": 10172.576665878296, |
| "prefill_time_avg_ms": 11.70957088470459, |
| "decode_time_avg_ms": 339.08588886260986, |
| "decode_tokens_per_s": 377.48548141990887, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 13.564988573960065, |
| "est_decode_energy_j": 786.2299462316408, |
| "est_total_infer_energy_j": 799.7949348056009, |
| "est_j_per_output_token": 0.20827993093895855 |
| }, |
| { |
| "axis": "01_model_scaling_initial", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/01_model_scaling_initial/llm_llama3.1-8b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "run_name": "llm_llama3.1-8b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "model": "llama3.1-8b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.019993, |
| "inference_energy_j": 0.175954, |
| "inference_power_avg_w": 8.801, |
| "inference_power_peak_w": 8.801, |
| "inference_prefill_duration_s": 0.749094, |
| "inference_prefill_energy_j": 39.663901, |
| "inference_prefill_power_avg_w": 52.983093, |
| "inference_prefill_power_peak_w": 80.049, |
| "inference_decode_duration_s": 50.320099, |
| "inference_decode_energy_j": 11396.476805, |
| "inference_decode_power_avg_w": 226.479781, |
| "inference_decode_power_peak_w": 435.624, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.030832, |
| "inference_idle_energy_j": 22033.793405, |
| "inference_idle_power_avg_w": 151.924723, |
| "inference_idle_power_peak_w": 435.54, |
| "idle_pre_duration_s": 30.010067, |
| "idle_pre_energy_j": 1603.038903, |
| "idle_pre_power_avg_w": 53.416613, |
| "idle_pre_power_peak_w": 80.009, |
| "idle_mid_duration_s": 32.500002, |
| "idle_mid_energy_j": 323.112906, |
| "idle_mid_power_avg_w": 9.941925, |
| "idle_mid_power_peak_w": 15.166, |
| "idle_post_duration_s": 29.989985, |
| "idle_post_energy_j": 2011.70938, |
| "idle_post_power_avg_w": 67.059371, |
| "idle_post_power_peak_w": 434.985, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 383.0, |
| "total_output_tokens": 3840.0, |
| "prefill_time_total_ms": 751.5397071838379, |
| "decode_time_total_ms": 50284.16037559509, |
| "prefill_time_avg_ms": 25.051323572794598, |
| "decode_time_avg_ms": 1676.1386791865032, |
| "decode_tokens_per_s": 76.3659961967607, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 39.81889819891405, |
| "est_decode_energy_j": 11388.345629633655, |
| "est_total_infer_energy_j": 11428.164527832569, |
| "est_j_per_output_token": 2.976084512456398 |
| }, |
| { |
| "axis": "01_model_scaling_initial", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/01_model_scaling_initial/llm_mistral-7b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "run_name": "llm_mistral-7b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "model": "mistral-7b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.019997, |
| "inference_energy_j": 0.166494, |
| "inference_power_avg_w": 8.326, |
| "inference_power_peak_w": 8.326, |
| "inference_prefill_duration_s": 0.770235, |
| "inference_prefill_energy_j": 39.353198, |
| "inference_prefill_power_avg_w": 51.087766, |
| "inference_prefill_power_peak_w": 78.997, |
| "inference_decode_duration_s": 44.859541, |
| "inference_decode_energy_j": 9944.680687, |
| "inference_decode_power_avg_w": 221.68444, |
| "inference_decode_power_peak_w": 428.021, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.050237, |
| "inference_idle_energy_j": 21224.019556, |
| "inference_idle_power_avg_w": 146.321186, |
| "inference_idle_power_peak_w": 428.021, |
| "idle_pre_duration_s": 30.010116, |
| "idle_pre_energy_j": 1623.648032, |
| "idle_pre_power_avg_w": 54.103225, |
| "idle_pre_power_peak_w": 79.894, |
| "idle_mid_duration_s": 32.47995, |
| "idle_mid_energy_j": 319.445465, |
| "idle_mid_power_avg_w": 9.835167, |
| "idle_mid_power_peak_w": 15.212, |
| "idle_post_duration_s": 29.989994, |
| "idle_post_energy_j": 1991.507512, |
| "idle_post_power_avg_w": 66.38616, |
| "idle_post_power_peak_w": 427.244, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 430.0, |
| "total_output_tokens": 3562.0, |
| "prefill_time_total_ms": 776.4148712158203, |
| "decode_time_total_ms": 44843.59335899353, |
| "prefill_time_avg_ms": 25.88049570719401, |
| "decode_time_avg_ms": 1494.7864452997844, |
| "decode_tokens_per_s": 79.43163634288975, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 39.66530125959397, |
| "est_decode_energy_j": 9941.1268813762, |
| "est_total_infer_energy_j": 9980.792182635794, |
| "est_j_per_output_token": 2.8020191416720364 |
| }, |
| { |
| "axis": "01_model_scaling_initial", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/01_model_scaling_initial/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.019997, |
| "inference_energy_j": 0.181133, |
| "inference_power_avg_w": 9.058, |
| "inference_power_peak_w": 9.058, |
| "inference_prefill_duration_s": 1.18955, |
| "inference_prefill_energy_j": 52.329583, |
| "inference_prefill_power_avg_w": 44.006042, |
| "inference_prefill_power_peak_w": 78.961, |
| "inference_decode_duration_s": 52.330052, |
| "inference_decode_energy_j": 9071.174013, |
| "inference_decode_power_avg_w": 173.34491, |
| "inference_decode_power_peak_w": 298.237, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.030406, |
| "inference_idle_energy_j": 18406.707336, |
| "inference_idle_power_avg_w": 126.916123, |
| "inference_idle_power_peak_w": 298.645, |
| "idle_pre_duration_s": 30.010068, |
| "idle_pre_energy_j": 1642.108328, |
| "idle_pre_power_avg_w": 54.718406, |
| "idle_pre_power_peak_w": 80.824, |
| "idle_mid_duration_s": 32.559997, |
| "idle_mid_energy_j": 332.204899, |
| "idle_mid_power_avg_w": 10.202868, |
| "idle_mid_power_peak_w": 15.328, |
| "idle_post_duration_s": 29.990067, |
| "idle_post_energy_j": 1942.928666, |
| "idle_post_power_avg_w": 64.766899, |
| "idle_post_power_peak_w": 293.197, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 1154.5767784118652, |
| "decode_time_total_ms": 52338.07301521301, |
| "prefill_time_avg_ms": 38.48589261372884, |
| "decode_time_avg_ms": 1744.6024338404338, |
| "decode_tokens_per_s": 72.85327449659222, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 50.80835420301724, |
| "est_decode_energy_j": 9072.538556395528, |
| "est_total_infer_energy_j": 9123.346910598546, |
| "est_j_per_output_token": 2.392695229635076 |
| }, |
| { |
| "axis": "02_decode_length_32_128", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/02_decode_length_32_128/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.010105, |
| "inference_energy_j": 0.083383, |
| "inference_power_avg_w": 8.252, |
| "inference_power_peak_w": 8.252, |
| "inference_prefill_duration_s": 0.859289, |
| "inference_prefill_energy_j": 46.769523, |
| "inference_prefill_power_avg_w": 54.453512, |
| "inference_prefill_power_peak_w": 78.799, |
| "inference_decode_duration_s": 50.449923, |
| "inference_decode_energy_j": 8827.995613, |
| "inference_decode_power_avg_w": 174.985012, |
| "inference_decode_power_peak_w": 307.146, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.040859, |
| "inference_idle_energy_j": 18347.689601, |
| "inference_idle_power_avg_w": 126.5004, |
| "inference_idle_power_peak_w": 307.146, |
| "idle_pre_duration_s": 30.00013, |
| "idle_pre_energy_j": 1604.198586, |
| "idle_pre_power_avg_w": 53.472922, |
| "idle_pre_power_peak_w": 81.719, |
| "idle_mid_duration_s": 32.49993, |
| "idle_mid_energy_j": 306.339518, |
| "idle_mid_power_avg_w": 9.425852, |
| "idle_mid_power_peak_w": 14.735, |
| "idle_post_duration_s": 29.990028, |
| "idle_post_energy_j": 1829.969002, |
| "idle_post_power_avg_w": 61.001455, |
| "idle_post_power_peak_w": 299.566, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 823.1363296508789, |
| "decode_time_total_ms": 50465.76762199402, |
| "prefill_time_avg_ms": 27.437877655029297, |
| "decode_time_avg_ms": 1682.1922540664673, |
| "decode_tokens_per_s": 75.55616766915513, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 44.82266400428009, |
| "est_decode_energy_j": 8830.752952923835, |
| "est_total_infer_energy_j": 8875.575616928114, |
| "est_j_per_output_token": 2.3277145599077143 |
| }, |
| { |
| "axis": "02_decode_length_32_128", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/02_decode_length_32_128/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok32_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok32_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 32, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.019997, |
| "inference_energy_j": 0.171331, |
| "inference_power_avg_w": 8.568, |
| "inference_power_peak_w": 8.568, |
| "inference_prefill_duration_s": 0.799165, |
| "inference_prefill_energy_j": 42.153896, |
| "inference_prefill_power_avg_w": 52.7747, |
| "inference_prefill_power_peak_w": 78.645, |
| "inference_decode_duration_s": 12.579192, |
| "inference_decode_energy_j": 1071.341996, |
| "inference_decode_power_avg_w": 85.168576, |
| "inference_decode_power_peak_w": 142.566, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.031649, |
| "inference_idle_energy_j": 15060.587644, |
| "inference_idle_power_avg_w": 103.843522, |
| "inference_idle_power_peak_w": 182.323, |
| "idle_pre_duration_s": 30.01007, |
| "idle_pre_energy_j": 1574.056232, |
| "idle_pre_power_avg_w": 52.450822, |
| "idle_pre_power_peak_w": 80.266, |
| "idle_mid_duration_s": 32.589995, |
| "idle_mid_energy_j": 315.691709, |
| "idle_mid_power_avg_w": 9.686778, |
| "idle_mid_power_peak_w": 14.891, |
| "idle_post_duration_s": 29.990004, |
| "idle_post_energy_j": 1738.090513, |
| "idle_post_power_avg_w": 57.939365, |
| "idle_post_power_peak_w": 155.408, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 960.0, |
| "prefill_time_total_ms": 813.4841918945312, |
| "decode_time_total_ms": 12541.171073913574, |
| "prefill_time_avg_ms": 27.116139729817707, |
| "decode_time_avg_ms": 418.03903579711914, |
| "decode_tokens_per_s": 76.54787534131167, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 42.93138418197632, |
| "est_decode_energy_j": 1068.1136817376098, |
| "est_total_infer_energy_j": 1111.045065919586, |
| "est_j_per_output_token": 1.1573386103329022 |
| }, |
| { |
| "axis": "03_dataset_alpaca_mmlu_longbench", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/03_dataset_alpaca_mmlu_longbench/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.012503, |
| "inference_energy_j": 0.107862, |
| "inference_power_avg_w": 8.627, |
| "inference_power_peak_w": 8.627, |
| "inference_prefill_duration_s": 0.780032, |
| "inference_prefill_energy_j": 40.855412, |
| "inference_prefill_power_avg_w": 52.386, |
| "inference_prefill_power_peak_w": 78.694, |
| "inference_decode_duration_s": 50.37973, |
| "inference_decode_energy_j": 8754.992437, |
| "inference_decode_power_avg_w": 173.779675, |
| "inference_decode_power_peak_w": 306.055, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.090316, |
| "inference_idle_energy_j": 18490.214541, |
| "inference_idle_power_avg_w": 127.439403, |
| "inference_idle_power_peak_w": 308.378, |
| "idle_pre_duration_s": 30.010114, |
| "idle_pre_energy_j": 1654.858879, |
| "idle_pre_power_avg_w": 55.143311, |
| "idle_pre_power_peak_w": 81.191, |
| "idle_mid_duration_s": 32.569917, |
| "idle_mid_energy_j": 328.669217, |
| "idle_mid_power_avg_w": 10.091156, |
| "idle_mid_power_peak_w": 14.976, |
| "idle_post_duration_s": 29.990025, |
| "idle_post_energy_j": 1896.990701, |
| "idle_post_power_avg_w": 63.237833, |
| "idle_post_power_peak_w": 301.287, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 803.7416934967041, |
| "decode_time_total_ms": 50382.55953788757, |
| "prefill_time_avg_ms": 26.79138978322347, |
| "decode_time_avg_ms": 1679.4186512629192, |
| "decode_tokens_per_s": 75.68095061015374, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 42.10481235551834, |
| "est_decode_energy_j": 8755.464822162252, |
| "est_total_infer_energy_j": 8797.569634517771, |
| "est_j_per_output_token": 2.307256657361073 |
| }, |
| { |
| "axis": "03_dataset_alpaca_mmlu_longbench", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/03_dataset_alpaca_mmlu_longbench/llm_qwen3-4b_gpu0_ds_longbench_fixed_maxtok128_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_longbench_fixed_maxtok128_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "longbench", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 4096, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.040708, |
| "inference_energy_j": 0.348621, |
| "inference_power_avg_w": 8.564, |
| "inference_power_peak_w": 8.564, |
| "inference_prefill_duration_s": 4.189805, |
| "inference_prefill_energy_j": 305.237196, |
| "inference_prefill_power_avg_w": 72.851988, |
| "inference_prefill_power_peak_w": 93.708, |
| "inference_decode_duration_s": 54.299759, |
| "inference_decode_energy_j": 12306.165997, |
| "inference_decode_power_avg_w": 226.633917, |
| "inference_decode_power_peak_w": 357.789, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.580408, |
| "inference_idle_energy_j": 18628.981504, |
| "inference_idle_power_avg_w": 127.963353, |
| "inference_idle_power_peak_w": 318.174, |
| "idle_pre_duration_s": 30.000164, |
| "idle_pre_energy_j": 1576.83161, |
| "idle_pre_power_avg_w": 52.560825, |
| "idle_pre_power_peak_w": 81.158, |
| "idle_mid_duration_s": 34.057265, |
| "idle_mid_energy_j": 333.623022, |
| "idle_mid_power_avg_w": 9.796193, |
| "idle_mid_power_peak_w": 14.857, |
| "idle_post_duration_s": 29.989995, |
| "idle_post_energy_j": 1898.624373, |
| "idle_post_power_avg_w": 63.29235, |
| "idle_post_power_peak_w": 312.159, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 122880.0, |
| "total_output_tokens": 3840.0, |
| "prefill_time_total_ms": 4171.168088912964, |
| "decode_time_total_ms": 54290.35472869873, |
| "prefill_time_avg_ms": 139.0389362970988, |
| "decode_time_avg_ms": 1809.6784909566243, |
| "decode_tokens_per_s": 70.73079590636965, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 303.8778875594702, |
| "est_decode_energy_j": 12304.035747484466, |
| "est_total_infer_energy_j": 12607.913635043937, |
| "est_j_per_output_token": 3.2833108424593584 |
| }, |
| { |
| "axis": "03_dataset_alpaca_mmlu_longbench", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/03_dataset_alpaca_mmlu_longbench/llm_qwen3-4b_gpu0_ds_mmlu-pro_fixed_maxtok128_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_mmlu-pro_fixed_maxtok128_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "mmlu-pro", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.029263, |
| "inference_energy_j": 0.411496, |
| "inference_power_avg_w": 14.062, |
| "inference_power_peak_w": 14.062, |
| "inference_prefill_duration_s": 0.950341, |
| "inference_prefill_energy_j": 57.671817, |
| "inference_prefill_power_avg_w": 60.681747, |
| "inference_prefill_power_peak_w": 79.125, |
| "inference_decode_duration_s": 53.989575, |
| "inference_decode_energy_j": 9909.655592, |
| "inference_decode_power_avg_w": 183.547178, |
| "inference_decode_power_peak_w": 294.889, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.07009, |
| "inference_idle_energy_j": 17868.819869, |
| "inference_idle_power_avg_w": 123.173523, |
| "inference_idle_power_peak_w": 294.889, |
| "idle_pre_duration_s": 30.010097, |
| "idle_pre_energy_j": 1591.493216, |
| "idle_pre_power_avg_w": 53.03189, |
| "idle_pre_power_peak_w": 81.269, |
| "idle_mid_duration_s": 32.53075, |
| "idle_mid_energy_j": 318.208575, |
| "idle_mid_power_avg_w": 9.781671, |
| "idle_mid_power_peak_w": 14.584, |
| "idle_post_duration_s": 29.990003, |
| "idle_post_energy_j": 1861.256906, |
| "idle_post_power_avg_w": 62.044558, |
| "idle_post_power_peak_w": 293.019, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 5055.0, |
| "total_output_tokens": 3840.0, |
| "prefill_time_total_ms": 933.809757232666, |
| "decode_time_total_ms": 53982.61785507202, |
| "prefill_time_avg_ms": 31.126991907755535, |
| "decode_time_avg_ms": 1799.4205951690674, |
| "decode_tokens_per_s": 71.13400854899827, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 56.66520743452406, |
| "est_decode_energy_j": 9908.357168350884, |
| "est_total_infer_energy_j": 9965.022375785407, |
| "est_j_per_output_token": 2.5950579103607834 |
| }, |
| { |
| "axis": "04_decode_length_32_256_512", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/04_decode_length_32_256_512/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok256_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok256_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 256, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.019992, |
| "inference_energy_j": 0.14908, |
| "inference_power_avg_w": 7.457, |
| "inference_power_peak_w": 7.457, |
| "inference_prefill_duration_s": 0.849989, |
| "inference_prefill_energy_j": 46.642753, |
| "inference_prefill_power_avg_w": 54.875129, |
| "inference_prefill_power_peak_w": 78.824, |
| "inference_decode_duration_s": 101.828956, |
| "inference_decode_energy_j": 23629.733066, |
| "inference_decode_power_avg_w": 232.052152, |
| "inference_decode_power_peak_w": 296.876, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.021068, |
| "inference_idle_energy_j": 17810.108217, |
| "inference_idle_power_avg_w": 122.809769, |
| "inference_idle_power_peak_w": 294.409, |
| "idle_pre_duration_s": 30.010071, |
| "idle_pre_energy_j": 1602.606878, |
| "idle_pre_power_avg_w": 53.402125, |
| "idle_pre_power_peak_w": 81.448, |
| "idle_mid_duration_s": 32.789999, |
| "idle_mid_energy_j": 305.828492, |
| "idle_mid_power_avg_w": 9.32686, |
| "idle_mid_power_peak_w": 15.038, |
| "idle_post_duration_s": 29.990056, |
| "idle_post_energy_j": 1844.239861, |
| "idle_post_power_avg_w": 61.476987, |
| "idle_post_power_peak_w": 288.339, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 7205.0, |
| "prefill_time_total_ms": 828.0079364776611, |
| "decode_time_total_ms": 101814.5387172699, |
| "prefill_time_avg_ms": 27.60026454925537, |
| "decode_time_avg_ms": 3393.81795724233, |
| "decode_tokens_per_s": 70.7659248941613, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 45.43704232723546, |
| "est_decode_energy_j": 23626.2828142298, |
| "est_total_infer_energy_j": 23671.719856557032, |
| "est_j_per_output_token": 3.285457301395841 |
| }, |
| { |
| "axis": "04_decode_length_32_256_512", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/04_decode_length_32_256_512/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok32_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok32_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 32, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.019993, |
| "inference_energy_j": 0.177717, |
| "inference_power_avg_w": 8.889, |
| "inference_power_peak_w": 8.889, |
| "inference_prefill_duration_s": 0.859627, |
| "inference_prefill_energy_j": 46.296709, |
| "inference_prefill_power_avg_w": 53.868209, |
| "inference_prefill_power_peak_w": 78.443, |
| "inference_decode_duration_s": 13.169953, |
| "inference_decode_energy_j": 1106.066293, |
| "inference_decode_power_avg_w": 83.984027, |
| "inference_decode_power_peak_w": 163.969, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.060436, |
| "inference_idle_energy_j": 15055.021566, |
| "inference_idle_power_avg_w": 103.784482, |
| "inference_idle_power_peak_w": 181.39, |
| "idle_pre_duration_s": 30.010115, |
| "idle_pre_energy_j": 1585.73269, |
| "idle_pre_power_avg_w": 52.839938, |
| "idle_pre_power_peak_w": 81.103, |
| "idle_mid_duration_s": 32.599953, |
| "idle_mid_energy_j": 308.415283, |
| "idle_mid_power_avg_w": 9.460576, |
| "idle_mid_power_peak_w": 14.453, |
| "idle_post_duration_s": 29.990037, |
| "idle_post_energy_j": 1767.194584, |
| "idle_post_power_avg_w": 58.909365, |
| "idle_post_power_peak_w": 159.079, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 960.0, |
| "prefill_time_total_ms": 836.2362384796143, |
| "decode_time_total_ms": 13192.95358657837, |
| "prefill_time_avg_ms": 27.87454128265381, |
| "decode_time_avg_ms": 439.7651195526123, |
| "decode_tokens_per_s": 72.76611667736329, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 45.046548467793706, |
| "est_decode_energy_j": 1107.9973702249445, |
| "est_total_infer_energy_j": 1153.043918692738, |
| "est_j_per_output_token": 1.2010874153049356 |
| }, |
| { |
| "axis": "04_decode_length_32_256_512", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/04_decode_length_32_256_512/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok512_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok512_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 512, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.019999, |
| "inference_energy_j": 0.16929, |
| "inference_power_avg_w": 8.465, |
| "inference_power_peak_w": 8.465, |
| "inference_prefill_duration_s": 0.808914, |
| "inference_prefill_energy_j": 43.824297, |
| "inference_prefill_power_avg_w": 54.208815, |
| "inference_prefill_power_peak_w": 79.14, |
| "inference_decode_duration_s": 173.619887, |
| "inference_decode_energy_j": 46206.355583, |
| "inference_decode_power_avg_w": 266.134713, |
| "inference_decode_power_peak_w": 309.218, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.051243, |
| "inference_idle_energy_j": 18530.060449, |
| "inference_idle_power_avg_w": 127.748967, |
| "inference_idle_power_peak_w": 309.218, |
| "idle_pre_duration_s": 30.010091, |
| "idle_pre_energy_j": 1574.822336, |
| "idle_pre_power_avg_w": 52.476258, |
| "idle_pre_power_peak_w": 81.83, |
| "idle_mid_duration_s": 32.689973, |
| "idle_mid_energy_j": 313.902395, |
| "idle_mid_power_avg_w": 9.602183, |
| "idle_mid_power_peak_w": 14.884, |
| "idle_post_duration_s": 29.989958, |
| "idle_post_energy_j": 1903.923634, |
| "idle_post_power_avg_w": 63.467131, |
| "idle_post_power_peak_w": 306.079, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 13162.0, |
| "prefill_time_total_ms": 790.414571762085, |
| "decode_time_total_ms": 173633.0442428589, |
| "prefill_time_avg_ms": 26.3471523920695, |
| "decode_time_avg_ms": 5787.76814142863, |
| "decode_tokens_per_s": 75.8035433715626, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 42.84743729395509, |
| "est_decode_energy_j": 46209.780396889546, |
| "est_total_infer_energy_j": 46252.6278341835, |
| "est_j_per_output_token": 3.5141033151636147 |
| }, |
| { |
| "axis": "05_inference_pattern_fixed_var_burst", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/05_inference_pattern_fixed_var_burst/llm_qwen3-4b_gpu0_ds_alpaca_burst_maxtok128_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_burst_maxtok128_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "burst", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.019991, |
| "inference_energy_j": 0.149735, |
| "inference_power_avg_w": 7.49, |
| "inference_power_peak_w": 7.49, |
| "inference_prefill_duration_s": 0.879419, |
| "inference_prefill_energy_j": 77.123012, |
| "inference_prefill_power_avg_w": 87.741682, |
| "inference_prefill_power_peak_w": 209.446, |
| "inference_decode_duration_s": 70.260518, |
| "inference_decode_energy_j": 15053.596857, |
| "inference_decode_power_avg_w": 214.254485, |
| "inference_decode_power_peak_w": 306.608, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 135.000096, |
| "inference_idle_energy_j": 12583.836466, |
| "inference_idle_power_avg_w": 93.213549, |
| "inference_idle_power_peak_w": 306.265, |
| "idle_pre_duration_s": 30.010114, |
| "idle_pre_energy_j": 1563.549671, |
| "idle_pre_power_avg_w": 52.100654, |
| "idle_pre_power_peak_w": 78.054, |
| "idle_mid_duration_s": 34.459961, |
| "idle_mid_energy_j": 310.820416, |
| "idle_mid_power_avg_w": 9.019747, |
| "idle_mid_power_peak_w": 14.669, |
| "idle_post_duration_s": 29.990058, |
| "idle_post_energy_j": 1848.648322, |
| "idle_post_power_avg_w": 61.624371, |
| "idle_post_power_peak_w": 304.928, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 842.4417972564697, |
| "decode_time_total_ms": 50234.103202819824, |
| "prefill_time_avg_ms": 28.081393241882324, |
| "decode_time_avg_ms": 1674.4701067606609, |
| "decode_tokens_per_s": 75.90460975495154, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 73.91726027838564, |
| "est_decode_energy_j": 10762.881911157012, |
| "est_total_infer_energy_j": 10836.799171435398, |
| "est_j_per_output_token": 2.8420663969146074 |
| }, |
| { |
| "axis": "05_inference_pattern_fixed_var_burst", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/05_inference_pattern_fixed_var_burst/llm_qwen3-4b_gpu0_ds_alpaca_concurrent_burst_maxtok128_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_concurrent_burst_maxtok128_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "concurrent_burst", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.0, |
| "inference_energy_j": 0.0, |
| "inference_power_avg_w": 0.0, |
| "inference_power_peak_w": 0.0, |
| "inference_prefill_duration_s": 0.769007, |
| "inference_prefill_energy_j": 134.572444, |
| "inference_prefill_power_avg_w": 175.111338, |
| "inference_prefill_power_peak_w": 302.308, |
| "inference_decode_duration_s": 51.260831, |
| "inference_decode_energy_j": 14560.259993, |
| "inference_decode_power_avg_w": 284.042518, |
| "inference_decode_power_peak_w": 302.308, |
| "inference_queue_duration_s": 0.060016, |
| "inference_queue_energy_j": 6.276485, |
| "inference_queue_power_avg_w": 104.672, |
| "inference_queue_power_peak_w": 299.765, |
| "inference_idle_duration_s": 12.110192, |
| "inference_idle_energy_j": 1059.109758, |
| "inference_idle_power_avg_w": 87.455975, |
| "inference_idle_power_peak_w": 300.681, |
| "idle_pre_duration_s": 30.010117, |
| "idle_pre_energy_j": 1551.503898, |
| "idle_pre_power_avg_w": 51.699203, |
| "idle_pre_power_peak_w": 78.606, |
| "idle_mid_duration_s": 34.029951, |
| "idle_mid_energy_j": 313.646389, |
| "idle_mid_power_avg_w": 9.216716, |
| "idle_mid_power_peak_w": 14.493, |
| "idle_post_duration_s": 29.990076, |
| "idle_post_energy_j": 1855.718753, |
| "idle_post_power_avg_w": 61.860187, |
| "idle_post_power_peak_w": 297.889, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 742.9652214050293, |
| "decode_time_total_ms": 51271.57258987427, |
| "prefill_time_avg_ms": 24.765507380167644, |
| "decode_time_avg_ms": 1709.0524196624756, |
| "decode_tokens_per_s": 74.36869609014173, |
| "queue_wait_avg_ms": 15612.065561612448, |
| "e2e_latency_avg_ms": 17348.279865582783, |
| "est_prefill_energy_j": 130.1016340077009, |
| "est_decode_energy_j": 14563.306580247667, |
| "est_total_infer_energy_j": 14693.408214255367, |
| "est_j_per_output_token": 3.853503334449349 |
| }, |
| { |
| "axis": "05_inference_pattern_fixed_var_burst", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/05_inference_pattern_fixed_var_burst/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.019569, |
| "inference_energy_j": 0.155574, |
| "inference_power_avg_w": 7.95, |
| "inference_power_peak_w": 7.95, |
| "inference_prefill_duration_s": 0.869939, |
| "inference_prefill_energy_j": 46.206683, |
| "inference_prefill_power_avg_w": 53.117425, |
| "inference_prefill_power_peak_w": 78.514, |
| "inference_decode_duration_s": 50.219731, |
| "inference_decode_energy_j": 8726.656689, |
| "inference_decode_power_avg_w": 173.768944, |
| "inference_decode_power_peak_w": 306.801, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.030374, |
| "inference_idle_energy_j": 18290.014867, |
| "inference_idle_power_avg_w": 126.111589, |
| "inference_idle_power_peak_w": 306.801, |
| "idle_pre_duration_s": 30.010085, |
| "idle_pre_energy_j": 1539.710153, |
| "idle_pre_power_avg_w": 51.306368, |
| "idle_pre_power_peak_w": 78.65, |
| "idle_mid_duration_s": 34.124767, |
| "idle_mid_energy_j": 316.317988, |
| "idle_mid_power_avg_w": 9.268048, |
| "idle_mid_power_peak_w": 14.899, |
| "idle_post_duration_s": 29.989972, |
| "idle_post_energy_j": 1860.32329, |
| "idle_post_power_avg_w": 62.013786, |
| "idle_post_power_peak_w": 301.779, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 839.3146991729736, |
| "decode_time_total_ms": 50214.84303474426, |
| "prefill_time_avg_ms": 27.97715663909912, |
| "decode_time_avg_ms": 1673.828101158142, |
| "decode_tokens_per_s": 75.933723368641, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 44.582235584717985, |
| "est_decode_energy_j": 8725.780247273266, |
| "est_total_infer_energy_j": 8770.362482857983, |
| "est_j_per_output_token": 2.3001212910721174 |
| }, |
| { |
| "axis": "05_inference_pattern_fixed_var_burst", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/05_inference_pattern_fixed_var_burst/llm_qwen3-4b_gpu0_ds_alpaca_variable_maxtok128_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_variable_maxtok128_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "variable", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.019996, |
| "inference_energy_j": 0.15553, |
| "inference_power_avg_w": 7.778, |
| "inference_power_peak_w": 7.778, |
| "inference_prefill_duration_s": 0.859922, |
| "inference_prefill_energy_j": 48.004394, |
| "inference_prefill_power_avg_w": 55.826674, |
| "inference_prefill_power_peak_w": 120.794, |
| "inference_decode_duration_s": 52.179703, |
| "inference_decode_energy_j": 9180.784151, |
| "inference_decode_power_avg_w": 175.945055, |
| "inference_decode_power_peak_w": 299.051, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 154.910412, |
| "inference_idle_energy_j": 19137.468548, |
| "inference_idle_power_avg_w": 123.539041, |
| "inference_idle_power_peak_w": 300.718, |
| "idle_pre_duration_s": 30.010073, |
| "idle_pre_energy_j": 1530.29808, |
| "idle_pre_power_avg_w": 50.992755, |
| "idle_pre_power_peak_w": 79.454, |
| "idle_mid_duration_s": 34.119995, |
| "idle_mid_energy_j": 322.801023, |
| "idle_mid_power_avg_w": 9.460761, |
| "idle_mid_power_peak_w": 15.037, |
| "idle_post_duration_s": 29.989974, |
| "idle_post_energy_j": 1814.721751, |
| "idle_post_power_avg_w": 60.494338, |
| "idle_post_power_peak_w": 294.682, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 854.7377586364746, |
| "decode_time_total_ms": 52178.68447303772, |
| "prefill_time_avg_ms": 28.49125862121582, |
| "decode_time_avg_ms": 1739.2894824345906, |
| "decode_tokens_per_s": 73.07581704115769, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 47.71716620688915, |
| "est_decode_energy_j": 9180.581509436268, |
| "est_total_infer_energy_j": 9228.298675643156, |
| "est_j_per_output_token": 2.4202199516504477 |
| }, |
| { |
| "axis": "06_power_cap_345_460_575", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/06_power_cap_345_460_575/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_cap345W", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_cap345W", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": 345, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.019388, |
| "inference_energy_j": 0.272806, |
| "inference_power_avg_w": 14.071, |
| "inference_power_peak_w": 14.071, |
| "inference_prefill_duration_s": 0.588955, |
| "inference_prefill_energy_j": 44.601832, |
| "inference_prefill_power_avg_w": 75.734119, |
| "inference_prefill_power_peak_w": 78.555, |
| "inference_decode_duration_s": 52.469806, |
| "inference_decode_energy_j": 9493.974764, |
| "inference_decode_power_avg_w": 180.940888, |
| "inference_decode_power_peak_w": 297.532, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.021276, |
| "inference_idle_energy_j": 17954.84139, |
| "inference_idle_power_avg_w": 123.808138, |
| "inference_idle_power_peak_w": 297.532, |
| "idle_pre_duration_s": 30.000101, |
| "idle_pre_energy_j": 1556.146732, |
| "idle_pre_power_avg_w": 51.87134, |
| "idle_pre_power_peak_w": 80.471, |
| "idle_mid_duration_s": 32.840578, |
| "idle_mid_energy_j": 305.60755, |
| "idle_mid_power_avg_w": 9.3057, |
| "idle_mid_power_peak_w": 14.588, |
| "idle_post_duration_s": 29.99001, |
| "idle_post_energy_j": 1862.867845, |
| "idle_post_power_avg_w": 62.099339, |
| "idle_post_power_peak_w": 290.696, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 547.9934215545654, |
| "decode_time_total_ms": 52478.959798812866, |
| "prefill_time_avg_ms": 18.26644738515218, |
| "decode_time_avg_ms": 1749.298659960429, |
| "decode_tokens_per_s": 72.6576901413022, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 41.50179899923063, |
| "est_decode_energy_j": 9495.589587313501, |
| "est_total_infer_energy_j": 9537.091386312732, |
| "est_j_per_output_token": 2.501204140129224 |
| }, |
| { |
| "axis": "06_power_cap_345_460_575", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/06_power_cap_345_460_575/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_cap460W", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_cap460W", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": 460, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.018789, |
| "inference_energy_j": 0.158957, |
| "inference_power_avg_w": 8.46, |
| "inference_power_peak_w": 8.46, |
| "inference_prefill_duration_s": 0.579816, |
| "inference_prefill_energy_j": 43.802933, |
| "inference_prefill_power_avg_w": 75.547138, |
| "inference_prefill_power_peak_w": 78.496, |
| "inference_decode_duration_s": 52.949512, |
| "inference_decode_energy_j": 9177.02313, |
| "inference_decode_power_avg_w": 173.31608, |
| "inference_decode_power_peak_w": 295.734, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.03068, |
| "inference_idle_energy_j": 18273.87197, |
| "inference_idle_power_avg_w": 125.9999, |
| "inference_idle_power_peak_w": 295.734, |
| "idle_pre_duration_s": 30.000125, |
| "idle_pre_energy_j": 1528.548729, |
| "idle_pre_power_avg_w": 50.951346, |
| "idle_pre_power_peak_w": 79.65, |
| "idle_mid_duration_s": 32.791145, |
| "idle_mid_energy_j": 323.683782, |
| "idle_mid_power_avg_w": 9.871129, |
| "idle_mid_power_peak_w": 14.78, |
| "idle_post_duration_s": 29.990001, |
| "idle_post_energy_j": 1859.518123, |
| "idle_post_power_avg_w": 61.986204, |
| "idle_post_power_peak_w": 290.877, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 551.727294921875, |
| "decode_time_total_ms": 52951.85351371765, |
| "prefill_time_avg_ms": 18.390909830729168, |
| "decode_time_avg_ms": 1765.0617837905884, |
| "decode_tokens_per_s": 72.00881077774186, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 41.68141808782959, |
| "est_decode_energy_j": 9177.40767973177, |
| "est_total_infer_energy_j": 9219.0890978196, |
| "est_j_per_output_token": 2.417804641442329 |
| }, |
| { |
| "axis": "06_power_cap_345_460_575", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/06_power_cap_345_460_575/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_cap575W", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_cap575W", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": 575, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.011994, |
| "inference_energy_j": 0.097563, |
| "inference_power_avg_w": 8.134, |
| "inference_power_peak_w": 8.134, |
| "inference_prefill_duration_s": 0.859622, |
| "inference_prefill_energy_j": 47.326467, |
| "inference_prefill_power_avg_w": 55.07064, |
| "inference_prefill_power_peak_w": 78.682, |
| "inference_decode_duration_s": 52.460004, |
| "inference_decode_energy_j": 9199.808958, |
| "inference_decode_power_avg_w": 175.368549, |
| "inference_decode_power_peak_w": 297.905, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.010446, |
| "inference_idle_energy_j": 18160.640822, |
| "inference_idle_power_avg_w": 125.236645, |
| "inference_idle_power_peak_w": 297.905, |
| "idle_pre_duration_s": 30.010091, |
| "idle_pre_energy_j": 1568.33057, |
| "idle_pre_power_avg_w": 52.260064, |
| "idle_pre_power_peak_w": 78.395, |
| "idle_mid_duration_s": 36.569974, |
| "idle_mid_energy_j": 341.400583, |
| "idle_mid_power_avg_w": 9.335538, |
| "idle_mid_power_peak_w": 15.08, |
| "idle_post_duration_s": 29.990009, |
| "idle_post_energy_j": 1881.640741, |
| "idle_post_power_avg_w": 62.72379, |
| "idle_post_power_peak_w": 295.481, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 821.3622570037842, |
| "decode_time_total_ms": 52446.71034812927, |
| "prefill_time_avg_ms": 27.37874190012614, |
| "decode_time_avg_ms": 1748.2236782709758, |
| "decode_tokens_per_s": 72.70236731131806, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 45.23294516504288, |
| "est_decode_energy_j": 9197.503493574715, |
| "est_total_infer_energy_j": 9242.736438739757, |
| "est_j_per_output_token": 2.4240064093206812 |
| }, |
| { |
| "axis": "07_clock_lock_1005_1500_2100", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/07_clock_lock_1005_1500_2100/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap_clk1005MHz", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap_clk1005MHz", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": 1005, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.009997, |
| "inference_energy_j": 0.521076, |
| "inference_power_avg_w": 52.123, |
| "inference_power_peak_w": 52.123, |
| "inference_prefill_duration_s": 0.919616, |
| "inference_prefill_energy_j": 48.198604, |
| "inference_prefill_power_avg_w": 52.411717, |
| "inference_prefill_power_peak_w": 52.966, |
| "inference_decode_duration_s": 68.839925, |
| "inference_decode_energy_j": 7431.261904, |
| "inference_decode_power_avg_w": 107.949798, |
| "inference_decode_power_peak_w": 146.07, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.0405, |
| "inference_idle_energy_j": 10041.985385, |
| "inference_idle_power_avg_w": 69.235787, |
| "inference_idle_power_peak_w": 145.668, |
| "idle_pre_duration_s": 30.010059, |
| "idle_pre_energy_j": 1578.600188, |
| "idle_pre_power_avg_w": 52.602366, |
| "idle_pre_power_peak_w": 53.281, |
| "idle_mid_duration_s": 34.320006, |
| "idle_mid_energy_j": 1790.617321, |
| "idle_mid_power_avg_w": 52.174156, |
| "idle_mid_power_peak_w": 52.579, |
| "idle_post_duration_s": 29.989967, |
| "idle_post_energy_j": 1659.949951, |
| "idle_post_power_avg_w": 55.349373, |
| "idle_post_power_peak_w": 145.487, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 977.8149127960205, |
| "decode_time_total_ms": 68757.69400596619, |
| "prefill_time_avg_ms": 32.59383042653402, |
| "decode_time_avg_ms": 2291.923133532206, |
| "decode_tokens_per_s": 55.45561198822551, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 51.24895848784471, |
| "est_decode_energy_j": 7422.379178889861, |
| "est_total_infer_energy_j": 7473.628137377706, |
| "est_j_per_output_token": 1.9600388506104656 |
| }, |
| { |
| "axis": "07_clock_lock_1005_1500_2100", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/07_clock_lock_1005_1500_2100/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap_clk1500MHz", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap_clk1500MHz", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": 1500, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.009997, |
| "inference_energy_j": 0.575004, |
| "inference_power_avg_w": 57.517, |
| "inference_power_peak_w": 57.517, |
| "inference_prefill_duration_s": 0.869312, |
| "inference_prefill_energy_j": 50.059388, |
| "inference_prefill_power_avg_w": 57.585333, |
| "inference_prefill_power_peak_w": 58.045, |
| "inference_decode_duration_s": 55.369519, |
| "inference_decode_energy_j": 6597.574192, |
| "inference_decode_power_avg_w": 119.154618, |
| "inference_decode_power_peak_w": 176.584, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.031211, |
| "inference_idle_energy_j": 11265.369284, |
| "inference_idle_power_avg_w": 77.675608, |
| "inference_idle_power_peak_w": 176.584, |
| "idle_pre_duration_s": 30.000118, |
| "idle_pre_energy_j": 1722.968228, |
| "idle_pre_power_avg_w": 57.432048, |
| "idle_pre_power_peak_w": 57.847, |
| "idle_mid_duration_s": 34.109948, |
| "idle_mid_energy_j": 1964.290672, |
| "idle_mid_power_avg_w": 57.587034, |
| "idle_mid_power_peak_w": 57.914, |
| "idle_post_duration_s": 29.990024, |
| "idle_post_energy_j": 1826.603277, |
| "idle_post_power_avg_w": 60.90587, |
| "idle_post_power_peak_w": 175.746, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 845.8425998687744, |
| "decode_time_total_ms": 55362.892627716064, |
| "prefill_time_avg_ms": 28.19475332895915, |
| "decode_time_avg_ms": 1845.4297542572021, |
| "decode_tokens_per_s": 68.87284639623608, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 48.70812777902913, |
| "est_decode_energy_j": 6596.744322430523, |
| "est_total_infer_energy_j": 6645.452450209553, |
| "est_j_per_output_token": 1.742840925835183 |
| }, |
| { |
| "axis": "07_clock_lock_1005_1500_2100", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/07_clock_lock_1005_1500_2100/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap_clk2100MHz", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap_clk2100MHz", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": 2100, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.009995, |
| "inference_energy_j": 0.610375, |
| "inference_power_avg_w": 61.069, |
| "inference_power_peak_w": 61.069, |
| "inference_prefill_duration_s": 0.839501, |
| "inference_prefill_energy_j": 51.456975, |
| "inference_prefill_power_avg_w": 61.294774, |
| "inference_prefill_power_peak_w": 62.029, |
| "inference_decode_duration_s": 52.699755, |
| "inference_decode_energy_j": 6749.58544, |
| "inference_decode_power_avg_w": 128.076519, |
| "inference_decode_power_peak_w": 199.308, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.050791, |
| "inference_idle_energy_j": 12293.659414, |
| "inference_idle_power_avg_w": 84.753892, |
| "inference_idle_power_peak_w": 198.658, |
| "idle_pre_duration_s": 30.010125, |
| "idle_pre_energy_j": 1839.246958, |
| "idle_pre_power_avg_w": 61.287545, |
| "idle_pre_power_peak_w": 62.117, |
| "idle_mid_duration_s": 34.109937, |
| "idle_mid_energy_j": 2091.870153, |
| "idle_mid_power_avg_w": 61.327294, |
| "idle_mid_power_peak_w": 61.899, |
| "idle_post_duration_s": 29.989999, |
| "idle_post_energy_j": 1950.212173, |
| "idle_post_power_avg_w": 65.027651, |
| "idle_post_power_peak_w": 199.308, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 828.8140296936035, |
| "decode_time_total_ms": 52705.62291145325, |
| "prefill_time_avg_ms": 27.627134323120117, |
| "decode_time_avg_ms": 1756.8540970484416, |
| "decode_tokens_per_s": 72.34522218636775, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 50.801968638098714, |
| "est_decode_energy_j": 6750.352714225576, |
| "est_total_infer_energy_j": 6801.154682863675, |
| "est_j_per_output_token": 1.7836755003576383 |
| }, |
| { |
| "axis": "08_ramp_rate", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/08_ramp_rate/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap_ramp", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap_ramp", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": true, |
| "ramp_start_w": 360, |
| "ramp_end_w": 600, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.021026, |
| "inference_energy_j": 0.171149, |
| "inference_power_avg_w": 8.14, |
| "inference_power_peak_w": 8.14, |
| "inference_prefill_duration_s": 0.810095, |
| "inference_prefill_energy_j": 44.128852, |
| "inference_prefill_power_avg_w": 54.474198, |
| "inference_prefill_power_peak_w": 78.347, |
| "inference_decode_duration_s": 50.399809, |
| "inference_decode_energy_j": 8679.418371, |
| "inference_decode_power_avg_w": 172.21214, |
| "inference_decode_power_peak_w": 303.047, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.030112, |
| "inference_idle_energy_j": 18200.421752, |
| "inference_idle_power_avg_w": 125.493478, |
| "inference_idle_power_peak_w": 303.047, |
| "idle_pre_duration_s": 30.010079, |
| "idle_pre_energy_j": 1589.174731, |
| "idle_pre_power_avg_w": 52.95466, |
| "idle_pre_power_peak_w": 79.173, |
| "idle_mid_duration_s": 34.119992, |
| "idle_mid_energy_j": 321.83639, |
| "idle_mid_power_avg_w": 9.432481, |
| "idle_mid_power_peak_w": 14.777, |
| "idle_post_duration_s": 29.990008, |
| "idle_post_energy_j": 287.173569, |
| "idle_post_power_avg_w": 9.575995, |
| "idle_post_power_peak_w": 14.773, |
| "ramp_duration_s": 61.081415, |
| "ramp_energy_j": 2026.231249, |
| "ramp_power_avg_w": 33.173074, |
| "ramp_power_peak_w": 299.544, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 801.2950420379639, |
| "decode_time_total_ms": 50383.25309753418, |
| "prefill_time_avg_ms": 26.709834734598797, |
| "decode_time_avg_ms": 1679.441769917806, |
| "decode_tokens_per_s": 75.67990881054509, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 43.649904776394365, |
| "est_decode_energy_j": 8676.60783608799, |
| "est_total_infer_energy_j": 8720.257740864383, |
| "est_j_per_output_token": 2.286980787008755 |
| }, |
| { |
| "axis": "09_control_combo_cap_clock_ramp", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/09_control_combo_cap_clock_ramp/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_cap345W_clk1500MHz", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_cap345W_clk1500MHz", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": 345, |
| "clock_lock_gpu": 1500, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.01, |
| "inference_energy_j": 0.579597, |
| "inference_power_avg_w": 57.958, |
| "inference_power_peak_w": 57.958, |
| "inference_prefill_duration_s": 0.879839, |
| "inference_prefill_energy_j": 50.764397, |
| "inference_prefill_power_avg_w": 57.697307, |
| "inference_prefill_power_peak_w": 57.958, |
| "inference_decode_duration_s": 55.9497, |
| "inference_decode_energy_j": 6412.671149, |
| "inference_decode_power_avg_w": 114.614791, |
| "inference_decode_power_peak_w": 176.152, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.010459, |
| "inference_idle_energy_j": 11633.926189, |
| "inference_idle_power_avg_w": 80.227918, |
| "inference_idle_power_peak_w": 175.923, |
| "idle_pre_duration_s": 30.010111, |
| "idle_pre_energy_j": 1734.058735, |
| "idle_pre_power_avg_w": 57.782482, |
| "idle_pre_power_peak_w": 58.23, |
| "idle_mid_duration_s": 34.18995, |
| "idle_mid_energy_j": 1980.635075, |
| "idle_mid_power_avg_w": 57.930329, |
| "idle_mid_power_peak_w": 58.228, |
| "idle_post_duration_s": 29.990041, |
| "idle_post_energy_j": 1833.727505, |
| "idle_post_power_avg_w": 61.143234, |
| "idle_post_power_peak_w": 176.152, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 855.8876514434814, |
| "decode_time_total_ms": 55930.803298950195, |
| "prefill_time_avg_ms": 28.529588381449383, |
| "decode_time_avg_ms": 1864.3601099650066, |
| "decode_tokens_per_s": 68.17352469657035, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 49.38241258284354, |
| "est_decode_energy_j": 6410.497330571287, |
| "est_total_infer_energy_j": 6459.879743154131, |
| "est_j_per_output_token": 1.6941725001715529 |
| }, |
| { |
| "axis": "09_control_combo_cap_clock_ramp", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/09_control_combo_cap_clock_ramp/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_cap345W_ramp", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_cap345W_ramp", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": 345, |
| "clock_lock_gpu": null, |
| "ramp_enabled": true, |
| "ramp_start_w": 360, |
| "ramp_end_w": 600, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.019982, |
| "inference_energy_j": 0.161596, |
| "inference_power_avg_w": 8.087, |
| "inference_power_peak_w": 8.087, |
| "inference_prefill_duration_s": 0.829873, |
| "inference_prefill_energy_j": 44.208895, |
| "inference_prefill_power_avg_w": 53.275855, |
| "inference_prefill_power_peak_w": 78.404, |
| "inference_decode_duration_s": 53.229312, |
| "inference_decode_energy_j": 9403.245324, |
| "inference_decode_power_avg_w": 176.654812, |
| "inference_decode_power_peak_w": 296.431, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.08082, |
| "inference_idle_energy_j": 17931.032851, |
| "inference_idle_power_avg_w": 123.593541, |
| "inference_idle_power_peak_w": 296.431, |
| "idle_pre_duration_s": 30.010087, |
| "idle_pre_energy_j": 1577.973433, |
| "idle_pre_power_avg_w": 52.581385, |
| "idle_pre_power_peak_w": 78.781, |
| "idle_mid_duration_s": 34.130019, |
| "idle_mid_energy_j": 305.964973, |
| "idle_mid_power_avg_w": 8.964689, |
| "idle_mid_power_peak_w": 14.696, |
| "idle_post_duration_s": 29.990001, |
| "idle_post_energy_j": 294.275039, |
| "idle_post_power_avg_w": 9.814043, |
| "idle_post_power_peak_w": 16.179, |
| "ramp_duration_s": 61.060931, |
| "ramp_energy_j": 2108.672314, |
| "ramp_power_avg_w": 34.534143, |
| "ramp_power_peak_w": 290.908, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 836.9174003601074, |
| "decode_time_total_ms": 53241.010665893555, |
| "prefill_time_avg_ms": 27.897246678670246, |
| "decode_time_avg_ms": 1774.7003555297852, |
| "decode_tokens_per_s": 71.61772386192935, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 44.58749006856203, |
| "est_decode_energy_j": 9405.28072987342, |
| "est_total_infer_energy_j": 9449.868219941982, |
| "est_j_per_output_token": 2.4783289325837874 |
| }, |
| { |
| "axis": "10_precision_bf16_fp16_4bit", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/10_precision_bf16_fp16_4bit/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.021406, |
| "inference_energy_j": 0.166856, |
| "inference_power_avg_w": 7.795, |
| "inference_power_peak_w": 7.795, |
| "inference_prefill_duration_s": 0.838977, |
| "inference_prefill_energy_j": 45.070436, |
| "inference_prefill_power_avg_w": 53.757024, |
| "inference_prefill_power_peak_w": 78.546, |
| "inference_decode_duration_s": 51.699637, |
| "inference_decode_energy_j": 9407.161052, |
| "inference_decode_power_avg_w": 181.957143, |
| "inference_decode_power_peak_w": 302.718, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.031467, |
| "inference_idle_energy_j": 17889.413117, |
| "inference_idle_power_avg_w": 123.348982, |
| "inference_idle_power_peak_w": 302.718, |
| "idle_pre_duration_s": 30.010115, |
| "idle_pre_energy_j": 1523.229924, |
| "idle_pre_power_avg_w": 50.757176, |
| "idle_pre_power_peak_w": 78.099, |
| "idle_mid_duration_s": 34.701777, |
| "idle_mid_energy_j": 330.486979, |
| "idle_mid_power_avg_w": 9.523738, |
| "idle_mid_power_peak_w": 14.567, |
| "idle_post_duration_s": 29.990034, |
| "idle_post_energy_j": 1856.605643, |
| "idle_post_power_avg_w": 61.889527, |
| "idle_post_power_peak_w": 302.707, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 831.2935829162598, |
| "decode_time_total_ms": 51671.080112457275, |
| "prefill_time_avg_ms": 27.70978609720866, |
| "decode_time_avg_ms": 1722.3693370819092, |
| "decode_tokens_per_s": 73.79369642944103, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 44.68786908787537, |
| "est_decode_energy_j": 9401.922112986844, |
| "est_total_infer_energy_j": 9446.609982074719, |
| "est_j_per_output_token": 2.4774744248819087 |
| }, |
| { |
| "axis": "10_precision_bf16_fp16_4bit", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/10_precision_bf16_fp16_4bit/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap_4bit", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap_4bit", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": true, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.029204, |
| "inference_energy_j": 0.429207, |
| "inference_power_avg_w": 14.697, |
| "inference_power_peak_w": 14.697, |
| "inference_prefill_duration_s": 1.509013, |
| "inference_prefill_energy_j": 99.163106, |
| "inference_prefill_power_avg_w": 65.722914, |
| "inference_prefill_power_peak_w": 78.54, |
| "inference_decode_duration_s": 94.069489, |
| "inference_decode_energy_j": 13155.090505, |
| "inference_decode_power_avg_w": 139.844313, |
| "inference_decode_power_peak_w": 168.835, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.061537, |
| "inference_idle_energy_j": 13718.14391, |
| "inference_idle_power_avg_w": 94.567774, |
| "inference_idle_power_peak_w": 168.695, |
| "idle_pre_duration_s": 30.010067, |
| "idle_pre_energy_j": 1577.487392, |
| "idle_pre_power_avg_w": 52.565105, |
| "idle_pre_power_peak_w": 87.534, |
| "idle_mid_duration_s": 34.383948, |
| "idle_mid_energy_j": 331.786146, |
| "idle_mid_power_avg_w": 9.655086, |
| "idle_mid_power_peak_w": 14.832, |
| "idle_post_duration_s": 29.989968, |
| "idle_post_energy_j": 1716.870926, |
| "idle_post_power_avg_w": 57.231869, |
| "idle_post_power_peak_w": 167.717, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3798.0, |
| "prefill_time_total_ms": 1472.3632335662842, |
| "decode_time_total_ms": 94113.22522163391, |
| "prefill_time_avg_ms": 49.07877445220947, |
| "decode_time_avg_ms": 3137.107507387797, |
| "decode_tokens_per_s": 40.35564598977264, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 96.76800217643881, |
| "est_decode_energy_j": 13161.199325333668, |
| "est_total_infer_energy_j": 13257.967327510107, |
| "est_j_per_output_token": 3.4907760209347307 |
| }, |
| { |
| "axis": "10_precision_bf16_fp16_4bit", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/10_precision_bf16_fp16_4bit/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap_fp16", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap_fp16", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": false, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.019879, |
| "inference_energy_j": 0.182887, |
| "inference_power_avg_w": 9.2, |
| "inference_power_peak_w": 9.2, |
| "inference_prefill_duration_s": 1.298783, |
| "inference_prefill_energy_j": 61.167749, |
| "inference_prefill_power_avg_w": 47.126131, |
| "inference_prefill_power_peak_w": 79.035, |
| "inference_decode_duration_s": 53.639719, |
| "inference_decode_energy_j": 9655.007355, |
| "inference_decode_power_avg_w": 179.996655, |
| "inference_decode_power_peak_w": 295.716, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.051536, |
| "inference_idle_energy_j": 18008.145089, |
| "inference_idle_power_avg_w": 124.15045, |
| "inference_idle_power_peak_w": 295.716, |
| "idle_pre_duration_s": 30.000163, |
| "idle_pre_energy_j": 1572.969553, |
| "idle_pre_power_avg_w": 52.432042, |
| "idle_pre_power_peak_w": 78.298, |
| "idle_mid_duration_s": 34.130021, |
| "idle_mid_energy_j": 332.306021, |
| "idle_mid_power_avg_w": 9.736483, |
| "idle_mid_power_peak_w": 15.194, |
| "idle_post_duration_s": 29.990015, |
| "idle_post_energy_j": 1852.697079, |
| "idle_post_power_avg_w": 61.75935, |
| "idle_post_power_peak_w": 292.59, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3840.0, |
| "prefill_time_total_ms": 1297.250509262085, |
| "decode_time_total_ms": 53627.22826004028, |
| "prefill_time_avg_ms": 43.241683642069496, |
| "decode_time_avg_ms": 1787.574275334676, |
| "decode_tokens_per_s": 71.60541621468309, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 61.13439743930173, |
| "est_decode_energy_j": 9652.721703728721, |
| "est_total_infer_energy_j": 9713.856101168023, |
| "est_j_per_output_token": 2.5296500263458395 |
| }, |
| { |
| "axis": "11_train_infer_split", |
| "scenario": "full_cycle", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/11_train_infer_split/full_cycle/llm_mistral-7b_gpu0_ds_longbench_fixed_maxtok256_nocap", |
| "run_name": "llm_mistral-7b_gpu0_ds_longbench_fixed_maxtok256_nocap", |
| "model": "mistral-7b", |
| "prompt_set": "longbench", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 256, |
| "max_seq_length": 4096, |
| "train_max_seq_length": 512, |
| "batch_size": 1, |
| "train_epochs": 3, |
| "train_max_steps": 200, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": false, |
| "skip_inference": false, |
| "train_duration_s": 84.705885, |
| "train_energy_j": 44473.933926, |
| "train_power_avg_w": 525.019738, |
| "train_power_peak_w": 537.105, |
| "validation_duration_s": 0.040064, |
| "validation_energy_j": 21.480802, |
| "validation_power_avg_w": 536.159, |
| "validation_power_peak_w": 536.159, |
| "inference_duration_s": 0.01766, |
| "inference_energy_j": 0.264384, |
| "inference_power_avg_w": 14.971, |
| "inference_power_peak_w": 14.971, |
| "inference_prefill_duration_s": 9.039371, |
| "inference_prefill_energy_j": 755.037893, |
| "inference_prefill_power_avg_w": 83.526908, |
| "inference_prefill_power_peak_w": 170.734, |
| "inference_decode_duration_s": 152.83049, |
| "inference_decode_energy_j": 47158.775225, |
| "inference_decode_power_avg_w": 308.569399, |
| "inference_decode_power_peak_w": 409.056, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.470177, |
| "inference_idle_energy_j": 19335.4513, |
| "inference_idle_power_avg_w": 132.917887, |
| "inference_idle_power_peak_w": 333.496, |
| "idle_pre_duration_s": 36.214987, |
| "idle_pre_energy_j": 1944.726426, |
| "idle_pre_power_avg_w": 53.889434, |
| "idle_pre_power_peak_w": 79.104, |
| "idle_mid_duration_s": 37.098219, |
| "idle_mid_energy_j": 2248.834564, |
| "idle_mid_power_avg_w": 60.641458, |
| "idle_mid_power_peak_w": 536.159, |
| "idle_post_duration_s": 29.989969, |
| "idle_post_energy_j": 1969.557927, |
| "idle_post_power_avg_w": 65.655345, |
| "idle_post_power_peak_w": 330.394, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 122880.0, |
| "total_output_tokens": 7586.0, |
| "prefill_time_total_ms": 9037.776231765747, |
| "decode_time_total_ms": 152872.2219467163, |
| "prefill_time_avg_ms": 301.2592077255249, |
| "decode_time_avg_ms": 5095.74073155721, |
| "decode_tokens_per_s": 49.623142147067796, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 754.8975038352843, |
| "est_decode_energy_j": 47171.689649892855, |
| "est_total_infer_energy_j": 47926.58715372814, |
| "est_j_per_output_token": 6.317767882115494 |
| }, |
| { |
| "axis": "11_train_infer_split", |
| "scenario": "infer_only", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/11_train_infer_split/infer_only/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.01999, |
| "inference_energy_j": 0.165854, |
| "inference_power_avg_w": 8.297, |
| "inference_power_peak_w": 8.297, |
| "inference_prefill_duration_s": 0.899857, |
| "inference_prefill_energy_j": 48.859658, |
| "inference_prefill_power_avg_w": 54.302878, |
| "inference_prefill_power_peak_w": 79.285, |
| "inference_decode_duration_s": 53.579913, |
| "inference_decode_energy_j": 9539.49891, |
| "inference_decode_power_avg_w": 178.042265, |
| "inference_decode_power_peak_w": 304.018, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.080289, |
| "inference_idle_energy_j": 18303.174138, |
| "inference_idle_power_avg_w": 126.159574, |
| "inference_idle_power_peak_w": 304.018, |
| "idle_pre_duration_s": 30.010115, |
| "idle_pre_energy_j": 1557.268988, |
| "idle_pre_power_avg_w": 51.891429, |
| "idle_pre_power_peak_w": 78.967, |
| "idle_mid_duration_s": 34.539949, |
| "idle_mid_energy_j": 341.042536, |
| "idle_mid_power_avg_w": 9.873882, |
| "idle_mid_power_peak_w": 14.817, |
| "idle_post_duration_s": 29.989958, |
| "idle_post_energy_j": 1947.29006, |
| "idle_post_power_avg_w": 64.912786, |
| "idle_post_power_peak_w": 292.821, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 887.5777721405029, |
| "decode_time_total_ms": 53607.79428482056, |
| "prefill_time_avg_ms": 29.585925738016766, |
| "decode_time_avg_ms": 1786.9264761606853, |
| "decode_tokens_per_s": 71.12771661041236, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 48.19802747605753, |
| "est_decode_energy_j": 9544.453116123506, |
| "est_total_infer_energy_j": 9592.651143599564, |
| "est_j_per_output_token": 2.5157752802516558 |
| }, |
| { |
| "axis": "11_train_infer_split", |
| "scenario": "train_only", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/11_train_infer_split/train_only/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": 200, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": false, |
| "skip_inference": true, |
| "train_duration_s": 220.944118, |
| "train_energy_j": 125718.553321, |
| "train_power_avg_w": 568.989625, |
| "train_power_peak_w": 591.943, |
| "validation_duration_s": 0.071788, |
| "validation_energy_j": 38.216202, |
| "validation_power_avg_w": 531.361429, |
| "validation_power_peak_w": 570.766, |
| "inference_duration_s": 0.0, |
| "inference_energy_j": 0.0, |
| "inference_power_avg_w": 0.0, |
| "inference_power_peak_w": 0.0, |
| "inference_prefill_duration_s": 0.0, |
| "inference_prefill_energy_j": 0.0, |
| "inference_prefill_power_avg_w": 0.0, |
| "inference_prefill_power_peak_w": 0.0, |
| "inference_decode_duration_s": 0.0, |
| "inference_decode_energy_j": 0.0, |
| "inference_decode_power_avg_w": 0.0, |
| "inference_decode_power_peak_w": 0.0, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 0.0, |
| "inference_idle_energy_j": 0.0, |
| "inference_idle_power_avg_w": 0.0, |
| "inference_idle_power_peak_w": 0.0, |
| "idle_pre_duration_s": 36.800562, |
| "idle_pre_energy_j": 2040.914519, |
| "idle_pre_power_avg_w": 55.534414, |
| "idle_pre_power_peak_w": 79.347, |
| "idle_mid_duration_s": 30.000044, |
| "idle_mid_energy_j": 2022.989904, |
| "idle_mid_power_avg_w": 67.432415, |
| "idle_mid_power_peak_w": 501.808, |
| "idle_post_duration_s": 29.989961, |
| "idle_post_energy_j": 300.621145, |
| "idle_post_power_avg_w": 10.025641, |
| "idle_post_power_peak_w": 15.324, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 0.0, |
| "total_input_tokens": 0.0, |
| "total_output_tokens": 0.0, |
| "prefill_time_total_ms": 0.0, |
| "decode_time_total_ms": 0.0, |
| "prefill_time_avg_ms": 0.0, |
| "decode_time_avg_ms": 0.0, |
| "decode_tokens_per_s": 0.0, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 0.0, |
| "est_decode_energy_j": 0.0, |
| "est_total_infer_energy_j": 0.0, |
| "est_j_per_output_token": 0.0 |
| }, |
| { |
| "axis": "12_model_scaling_tokenpowerbench", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/12_model_scaling_tokenpowerbench/llm_gpt2_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "run_name": "llm_gpt2_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "model": "gpt2", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.010002, |
| "inference_energy_j": 0.08441, |
| "inference_power_avg_w": 8.439, |
| "inference_power_peak_w": 8.439, |
| "inference_prefill_duration_s": 0.369506, |
| "inference_prefill_energy_j": 14.735179, |
| "inference_prefill_power_avg_w": 39.930649, |
| "inference_prefill_power_peak_w": 78.64, |
| "inference_decode_duration_s": 9.949569, |
| "inference_decode_energy_j": 771.366851, |
| "inference_decode_power_avg_w": 77.527618, |
| "inference_decode_power_peak_w": 97.273, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.030967, |
| "inference_idle_energy_j": 11965.029983, |
| "inference_idle_power_avg_w": 82.499881, |
| "inference_idle_power_peak_w": 100.49, |
| "idle_pre_duration_s": 30.010113, |
| "idle_pre_energy_j": 1639.224921, |
| "idle_pre_power_avg_w": 54.622347, |
| "idle_pre_power_peak_w": 79.597, |
| "idle_mid_duration_s": 34.439946, |
| "idle_mid_energy_j": 337.903279, |
| "idle_mid_power_avg_w": 9.811345, |
| "idle_mid_power_peak_w": 15.192, |
| "idle_post_duration_s": 29.99001, |
| "idle_post_energy_j": 1650.570857, |
| "idle_post_power_avg_w": 55.021901, |
| "idle_post_power_peak_w": 98.688, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 382.0, |
| "total_output_tokens": 3840.0, |
| "prefill_time_total_ms": 355.3440570831299, |
| "decode_time_total_ms": 9934.163093566895, |
| "prefill_time_avg_ms": 11.844801902770996, |
| "decode_time_avg_ms": 331.13876978556317, |
| "decode_tokens_per_s": 386.5448919886049, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 14.189118817622424, |
| "est_decode_energy_j": 770.1720014677525, |
| "est_total_infer_energy_j": 784.3611202853749, |
| "est_j_per_output_token": 0.2042607084076497 |
| }, |
| { |
| "axis": "12_model_scaling_tokenpowerbench", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/12_model_scaling_tokenpowerbench/llm_llama3.1-8b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "run_name": "llm_llama3.1-8b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "model": "llama3.1-8b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.019993, |
| "inference_energy_j": 0.164466, |
| "inference_power_avg_w": 8.226, |
| "inference_power_peak_w": 8.226, |
| "inference_prefill_duration_s": 0.809462, |
| "inference_prefill_energy_j": 43.917828, |
| "inference_prefill_power_avg_w": 54.272988, |
| "inference_prefill_power_peak_w": 78.986, |
| "inference_decode_duration_s": 50.22965, |
| "inference_decode_energy_j": 12342.91572, |
| "inference_decode_power_avg_w": 245.728081, |
| "inference_decode_power_peak_w": 434.973, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.030931, |
| "inference_idle_energy_j": 21417.123216, |
| "inference_idle_power_avg_w": 147.673086, |
| "inference_idle_power_peak_w": 434.973, |
| "idle_pre_duration_s": 30.010066, |
| "idle_pre_energy_j": 1620.091145, |
| "idle_pre_power_avg_w": 53.984804, |
| "idle_pre_power_peak_w": 79.034, |
| "idle_mid_duration_s": 34.502228, |
| "idle_mid_energy_j": 334.030762, |
| "idle_mid_power_avg_w": 9.681508, |
| "idle_mid_power_peak_w": 14.638, |
| "idle_post_duration_s": 29.990001, |
| "idle_post_energy_j": 1983.263039, |
| "idle_post_power_avg_w": 66.11291, |
| "idle_post_power_peak_w": 434.345, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 383.0, |
| "total_output_tokens": 3840.0, |
| "prefill_time_total_ms": 759.7558498382568, |
| "decode_time_total_ms": 50253.07083129883, |
| "prefill_time_avg_ms": 25.325194994608562, |
| "decode_time_avg_ms": 1675.1023610432942, |
| "decode_tokens_per_s": 76.41324075280899, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 41.23422012120152, |
| "est_decode_energy_j": 12348.590659732135, |
| "est_total_infer_energy_j": 12389.824879853337, |
| "est_j_per_output_token": 3.22651689579514 |
| }, |
| { |
| "axis": "12_model_scaling_tokenpowerbench", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/12_model_scaling_tokenpowerbench/llm_mistral-7b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "run_name": "llm_mistral-7b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "model": "mistral-7b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.018704, |
| "inference_energy_j": 0.162875, |
| "inference_power_avg_w": 8.708, |
| "inference_power_peak_w": 8.708, |
| "inference_prefill_duration_s": 0.799509, |
| "inference_prefill_energy_j": 42.059777, |
| "inference_prefill_power_avg_w": 52.624725, |
| "inference_prefill_power_peak_w": 78.988, |
| "inference_decode_duration_s": 44.819886, |
| "inference_decode_energy_j": 9962.215376, |
| "inference_decode_power_avg_w": 222.271874, |
| "inference_decode_power_peak_w": 426.932, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.020655, |
| "inference_idle_energy_j": 21049.351545, |
| "inference_idle_power_avg_w": 145.146722, |
| "inference_idle_power_peak_w": 426.932, |
| "idle_pre_duration_s": 30.010115, |
| "idle_pre_energy_j": 1539.173396, |
| "idle_pre_power_avg_w": 51.288399, |
| "idle_pre_power_peak_w": 79.112, |
| "idle_mid_duration_s": 34.141242, |
| "idle_mid_energy_j": 353.648816, |
| "idle_mid_power_avg_w": 10.358475, |
| "idle_mid_power_peak_w": 15.08, |
| "idle_post_duration_s": 29.990002, |
| "idle_post_energy_j": 1953.982724, |
| "idle_post_power_avg_w": 65.135946, |
| "idle_post_power_peak_w": 425.289, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 430.0, |
| "total_output_tokens": 3562.0, |
| "prefill_time_total_ms": 760.5900764465332, |
| "decode_time_total_ms": 44816.834926605225, |
| "prefill_time_avg_ms": 25.353002548217773, |
| "decode_time_avg_ms": 1493.8944975535076, |
| "decode_tokens_per_s": 79.47906196038493, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 40.02584361072778, |
| "est_decode_energy_j": 9961.521885885195, |
| "est_total_infer_energy_j": 10001.547729495922, |
| "est_j_per_output_token": 2.807846077904526 |
| }, |
| { |
| "axis": "12_model_scaling_tokenpowerbench", |
| "scenario": "", |
| "run_dir": "/home/hyunwoo/DataCenter_train/hyp_opt/outputs/llm_power_experiment/12_model_scaling_tokenpowerbench/llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "run_name": "llm_qwen3-4b_gpu0_ds_alpaca_fixed_maxtok128_nocap", |
| "model": "qwen3-4b", |
| "prompt_set": "alpaca", |
| "inference_pattern": "fixed", |
| "gen_max_new_tokens": 128, |
| "max_seq_length": 512, |
| "train_max_seq_length": null, |
| "batch_size": 4, |
| "train_epochs": 3, |
| "train_max_steps": null, |
| "use_4bit": false, |
| "use_bf16": true, |
| "power_cap": null, |
| "clock_lock_gpu": null, |
| "ramp_enabled": false, |
| "ramp_start_w": null, |
| "ramp_end_w": null, |
| "skip_training": true, |
| "skip_inference": false, |
| "train_duration_s": 0.0, |
| "train_energy_j": 0.0, |
| "train_power_avg_w": 0.0, |
| "train_power_peak_w": 0.0, |
| "validation_duration_s": 0.0, |
| "validation_energy_j": 0.0, |
| "validation_power_avg_w": 0.0, |
| "validation_power_peak_w": 0.0, |
| "inference_duration_s": 0.019993, |
| "inference_energy_j": 0.189977, |
| "inference_power_avg_w": 9.502, |
| "inference_power_peak_w": 9.502, |
| "inference_prefill_duration_s": 0.849124, |
| "inference_prefill_energy_j": 46.590352, |
| "inference_prefill_power_avg_w": 54.893588, |
| "inference_prefill_power_peak_w": 78.944, |
| "inference_decode_duration_s": 51.560046, |
| "inference_decode_energy_j": 8958.376365, |
| "inference_decode_power_avg_w": 173.745862, |
| "inference_decode_power_peak_w": 302.872, |
| "inference_queue_duration_s": 0.0, |
| "inference_queue_energy_j": 0.0, |
| "inference_queue_power_avg_w": 0.0, |
| "inference_queue_power_peak_w": 0.0, |
| "inference_idle_duration_s": 145.040873, |
| "inference_idle_energy_j": 18464.776332, |
| "inference_idle_power_avg_w": 127.307911, |
| "inference_idle_power_peak_w": 302.872, |
| "idle_pre_duration_s": 30.010074, |
| "idle_pre_energy_j": 1566.000293, |
| "idle_pre_power_avg_w": 52.182347, |
| "idle_pre_power_peak_w": 79.389, |
| "idle_mid_duration_s": 35.239994, |
| "idle_mid_energy_j": 347.417193, |
| "idle_mid_power_avg_w": 9.858594, |
| "idle_mid_power_peak_w": 15.411, |
| "idle_post_duration_s": 29.990013, |
| "idle_post_energy_j": 1871.902399, |
| "idle_post_power_avg_w": 62.400381, |
| "idle_post_power_peak_w": 301.192, |
| "ramp_duration_s": 0.0, |
| "ramp_energy_j": 0.0, |
| "ramp_power_avg_w": 0.0, |
| "ramp_power_peak_w": 0.0, |
| "n_requests": 30.0, |
| "total_input_tokens": 354.0, |
| "total_output_tokens": 3813.0, |
| "prefill_time_total_ms": 822.5698471069336, |
| "decode_time_total_ms": 51564.01252746582, |
| "prefill_time_avg_ms": 27.418994903564453, |
| "decode_time_avg_ms": 1718.800417582194, |
| "decode_tokens_per_s": 73.94692175999661, |
| "queue_wait_avg_ms": 0.0, |
| "e2e_latency_avg_ms": 0.0, |
| "est_prefill_energy_j": 45.15381028831101, |
| "est_decode_energy_j": 8959.033804763347, |
| "est_total_infer_energy_j": 9004.187615051658, |
| "est_j_per_output_token": 2.361444430907857 |
| } |
| ] |
