| { | |
| "base_model": "HuggingFaceTB/SmolVLM-256M-Instruct", | |
| "method": "layer_pruning_shortgpt_bi_score", | |
| "layers_removed": [ | |
| 11, | |
| 12, | |
| 13, | |
| 14 | |
| ], | |
| "n_layers_original": 30, | |
| "n_layers_remaining": 26, | |
| "params_before_M": 256.5, | |
| "params_after_M": 242.3, | |
| "param_reduction_pct": 5.5, | |
| "benchmarks": { | |
| "vqav2": { | |
| "accuracy": 0.0333, | |
| "avg_latency_s": 0.3427, | |
| "peak_memory_mb": 1891.1, | |
| "avg_memory_mb": 1881.5, | |
| "throughput_sps": 2.916, | |
| "avg_power_w": 107.9, | |
| "avg_gpu_util_pct": 16.6, | |
| "n_samples": 50, | |
| "n_evaluated": 50, | |
| "n_skipped": 0, | |
| "all_failed": false, | |
| "zero_accuracy_warning": false, | |
| "metrics": { | |
| "exact_match": 0.0333, | |
| "contains": 0.04, | |
| "token_f1": 0.04, | |
| "bleu": 0.04, | |
| "rouge_l": 0.04 | |
| } | |
| } | |
| } | |
| } |