There was no progress for a long time when pruning Qwen3-8B.

[damon-93]

./pyvenv-torch-pruning/bin/python app/prune_llm.py --model ./output/Qwen3-8B --pruning_ratio 0.5 --max_seq_len 4096 --save ./output/Qwen3-8B-pruned
torch 2.10.0
transformers 5.1.0
accelerate 1.12.0

of gpus: 4

loading llm model ./output/Qwen3-8B
torch_dtype is deprecated! Use dtype instead!
Loading weights: 100%|█| 399/399 [01:44<00:00, 3.84it/s, Materializing param=model.norm.weight
use device cuda:0
/home/peixiaojun/code/ModelSpeedUp/pyvenv-torch-pruning/lib/python3.13/site-packages/torch_pruning/dependency/graph.py:390: UserWarning: Unwrapped parameters detected: ['model.layers.2.input_layernorm.weight', 'model.layers.4.self_attn.k_norm.weight', 'model.layers.15.self_attn.q_norm.weight', 'model.layers.21.self_attn.q_norm.weight', 'model.layers.26.self_attn.k_norm.weight', 'model.layers.6.input_layernorm.weight', 'model.layers.14.post_attention_layernorm.weight', 'model.layers.16.input_layernorm.weight', 'model.layers.15.post_attention_layernorm.weight', 'model.layers.26.post_attention_layernorm.weight', 'model.layers.27.post_attention_layernorm.weight', 'model.layers.1.post_attention_layernorm.weight', 'model.layers.6.post_attention_layernorm.weight', 'model.layers.28.post_attention_layernorm.weight', 'model.layers.8.self_attn.q_norm.weight', 'model.layers.14.self_attn.k_norm.weight', 'model.layers.21.input_layernorm.weight', 'model.layers.33.input_layernorm.weight', 'model.layers.9.self_attn.q_norm.weight', 'model.layers.12.self_attn.q_norm.weight', 'model.layers.15.self_attn.k_norm.weight', 'model.layers.27.self_attn.k_norm.weight', 'model.layers.34.input_layernorm.weight', 'model.layers.9.post_attention_layernorm.weight', 'model.layers.16.self_attn.q_norm.weight', 'model.layers.18.self_attn.q_norm.weight', 'model.layers.28.self_attn.k_norm.weight', 'model.layers.29.self_attn.k_norm.weight', 'model.layers.7.self_attn.k_norm.weight', 'model.layers.10.self_attn.q_norm.weight', 'model.layers.16.post_attention_layernorm.weight', 'model.layers.19.input_layernorm.weight', 'model.layers.22.self_attn.q_norm.weight', 'model.layers.23.self_attn.q_norm.weight', 'model.layers.23.self_attn.k_norm.weight', 'model.layers.3.self_attn.q_norm.weight', 'model.layers.8.input_layernorm.weight', 'model.layers.9.input_layernorm.weight', 'model.layers.17.post_attention_layernorm.weight', 'model.layers.19.self_attn.k_norm.weight', 'model.layers.26.self_attn.q_norm.weight', 'model.layers.29.post_attention_layernorm.weight', 'model.layers.4.self_attn.q_norm.weight', 'model.layers.5.self_attn.q_norm.weight', 'model.layers.11.input_layernorm.weight', 'model.layers.18.post_attention_layernorm.weight', 'model.layers.22.input_layernorm.weight', 'model.layers.13.input_layernorm.weight', 'model.layers.16.self_attn.k_norm.weight', 'model.layers.24.input_layernorm.weight', 'model.layers.31.post_attention_layernorm.weight', 'model.layers.3.input_layernorm.weight', 'model.layers.6.self_attn.k_norm.weight', 'model.layers.11.self_attn.q_norm.weight', 'model.layers.17.self_attn.k_norm.weight', 'model.layers.23.input_layernorm.weight', 'model.layers.13.self_attn.q_norm.weight', 'model.layers.18.self_attn.k_norm.weight', 'model.layers.27.self_attn.q_norm.weight', 'model.layers.30.self_attn.k_norm.weight', 'model.layers.35.post_attention_layernorm.weight', 'model.layers.1.self_attn.k_norm.weight', 'model.layers.10.post_attention_layernorm.weight', 'model.layers.25.self_attn.q_norm.weight', 'model.layers.31.self_attn.k_norm.weight', 'model.layers.2.post_attention_layernorm.weight', 'model.layers.10.input_layernorm.weight', 'model.layers.1.input_layernorm.weight', 'model.layers.2.self_attn.q_norm.weight', 'model.layers.7.input_layernorm.weight', 'model.layers.20.post_attention_layernorm.weight', 'model.layers.21.post_attention_layernorm.weight', 'model.layers.28.self_attn.q_norm.weight', 'model.layers.29.self_attn.q_norm.weight', 'model.layers.32.post_attention_layernorm.weight', 'model.layers.34.self_attn.k_norm.weight', 'model.layers.0.post_attention_layernorm.weight', 'model.layers.19.post_attention_layernorm.weight', 'model.layers.25.input_layernorm.weight', 'model.layers.33.post_attention_layernorm.weight', 'model.norm.weight', 'model.layers.0.self_attn.q_norm.weight', 'model.layers.4.input_layernorm.weight', 'model.layers.8.self_attn.k_norm.weight', 'model.layers.34.post_attention_layernorm.weight', 'model.layers.0.input_layernorm.weight', 'model.layers.5.input_layernorm.weight', 'model.layers.14.self_attn.q_norm.weight', 'model.layers.32.self_attn.k_norm.weight', 'model.layers.8.post_attention_layernorm.weight', 'model.layers.33.self_attn.k_norm.weight', 'model.layers.12.input_layernorm.weight', 'model.layers.14.input_layernorm.weight', 'model.layers.30.self_attn.q_norm.weight', 'model.layers.23.post_attention_layernorm.weight', 'model.layers.26.input_layernorm.weight', 'model.layers.27.input_layernorm.weight', 'model.layers.31.self_attn.q_norm.weight', 'model.layers.32.self_attn.q_norm.weight', 'model.layers.17.self_attn.q_norm.weight', 'model.layers.3.post_attention_layernorm.weight', 'model.layers.5.self_attn.k_norm.weight', 'model.layers.9.self_attn.k_norm.weight', 'model.layers.20.self_attn.k_norm.weight', 'model.layers.21.self_attn.k_norm.weight', 'model.layers.0.self_attn.k_norm.weight', 'model.layers.4.post_attention_layernorm.weight', 'model.layers.7.post_attention_layernorm.weight', 'model.layers.10.self_attn.k_norm.weight', 'model.layers.29.input_layernorm.weight', 'model.layers.30.input_layernorm.weight', 'model.layers.1.self_attn.q_norm.weight', 'model.layers.24.self_attn.k_norm.weight', 'model.layers.35.self_attn.k_norm.weight', 'model.layers.3.self_attn.k_norm.weight', 'model.layers.11.post_attention_layernorm.weight', 'model.layers.19.self_attn.q_norm.weight', 'model.layers.12.post_attention_layernorm.weight', 'model.layers.15.input_layernorm.weight', 'model.layers.24.post_attention_layernorm.weight', 'model.layers.35.input_layernorm.weight', 'model.layers.13.post_attention_layernorm.weight', 'model.layers.17.input_layernorm.weight', 'model.layers.24.self_attn.q_norm.weight', 'model.layers.25.post_attention_layernorm.weight', 'model.layers.28.input_layernorm.weight', 'model.layers.33.self_attn.q_norm.weight', 'model.layers.34.self_attn.q_norm.weight', 'model.layers.11.self_attn.k_norm.weight', 'model.layers.18.input_layernorm.weight', 'model.layers.22.self_attn.k_norm.weight', 'model.layers.22.post_attention_layernorm.weight', 'model.layers.35.self_attn.q_norm.weight', 'model.layers.5.post_attention_layernorm.weight', 'model.layers.6.self_attn.q_norm.weight', 'model.layers.12.self_attn.k_norm.weight', 'model.layers.13.self_attn.k_norm.weight', 'model.layers.20.input_layernorm.weight', 'model.layers.30.post_attention_layernorm.weight', 'model.layers.31.input_layernorm.weight', 'model.layers.32.input_layernorm.weight', 'model.layers.2.self_attn.k_norm.weight', 'model.layers.7.self_attn.q_norm.weight', 'model.layers.20.self_attn.q_norm.weight', 'model.layers.25.self_attn.k_norm.weight'].
Torch-Pruning will prune the last non-singleton dimension of these parameters. If you wish to change this behavior, please provide an unwrapped_parameters argument.
warnings.warn(warning_str)

After startup, it gets stuck in the tp.pruner.BasePruner function.