FastVLM-0.5B-LAYER-PRUNED / pruning_info.json
Azaz666's picture
Upload layer-pruned FastVLM-0.5B (ShortGPT, removed 3 layers)
862b186 verified
{
"base_model": "apple/FastVLM-0.5B",
"method": "layer_pruning_shortgpt_bi_score",
"layers_removed": [
7,
8,
9
],
"n_layers_original": 24,
"n_layers_remaining": 21,
"params_before_M": 758.8,
"params_after_M": 714.1,
"param_reduction_pct": 5.9,
"benchmarks": {
"vqav2": {
"accuracy": 0.2933,
"avg_latency_s": 0.5645,
"peak_memory_mb": 2607.1,
"avg_memory_mb": 2564.0,
"throughput_sps": 1.77,
"avg_power_w": 102.8,
"avg_gpu_util_pct": 26.5,
"n_samples": 50,
"n_evaluated": 50,
"n_skipped": 0,
"all_failed": false,
"zero_accuracy_warning": false,
"metrics": {
"exact_match": 0.2933,
"contains": 0.44,
"token_f1": 0.3573,
"bleu": 0.3433,
"rouge_l": 0.3573
}
}
}
}