pruning_info.json

{
  "base_model": "LiquidAI/LFM2-VL-3B",
  "method": "layer_pruning_ppl",
  "method_name": "PPL-based (Shortened LLaMA)",
  "layers_removed": [
    3,
    6,
    7,
    10,
    11,
    14,
    15,
    18,
    25
  ],
  "n_layers_original": 30,
  "n_layers_remaining": 21,
  "params_before_M": 2999.0,
  "params_after_M": 2253.3,
  "param_reduction_pct": 24.9,
  "benchmarks": {
    "vqav2": {
      "accuracy": 0.0,
      "avg_latency_s": 0.0,
      "peak_memory_mb": 7981.1,
      "avg_memory_mb": 7981.0,
      "throughput_sps": 0.0,
      "avg_power_w": 235.4,
      "avg_gpu_util_pct": 57.2,
      "n_samples": 50,
      "n_evaluated": 0,
      "n_skipped": 50,
      "all_failed": true,
      "zero_accuracy_warning": false
    }
  }
}