pruning_info.json

{
  "base_model": "AIDC-AI/Ovis2-4B",
  "method": "layer_pruning_ppl",
  "method_name": "PPL-based (Shortened LLaMA)",
  "layers_removed": [
    3,
    4,
    5,
    6,
    7,
    8,
    10,
    11,
    18,
    19
  ],
  "n_layers_original": 36,
  "n_layers_remaining": 26,
  "params_before_M": 4304.9,
  "params_after_M": 3534.2,
  "param_reduction_pct": 17.9,
  "benchmarks": {
    "vqav2": {
      "accuracy": 0.06,
      "avg_latency_s": 0.5029,
      "peak_memory_mb": 15235.1,
      "avg_memory_mb": 15230.8,
      "throughput_sps": 1.987,
      "avg_power_w": 236.7,
      "avg_gpu_util_pct": 69.1,
      "n_samples": 50,
      "n_evaluated": 50,
      "n_skipped": 0,
      "all_failed": false,
      "zero_accuracy_warning": false,
      "metrics": {
        "exact_match": 0.06,
        "contains": 0.42,
        "token_f1": 0.1762,
        "bleu": 0.135,
        "rouge_l": 0.1762
      }
    }
  }
}