Verified-Intelligence / alpha-beta-CROWN

alpha-beta-CROWN: An Efficient, Scalable and GPU Accelerated Neural Network Verifier (winner of VNN-COMP 2021, 2022, 2023, and 2024)
Other
248 stars 60 forks source link

AttributeError: 'Patches' object has no attribute 'permute' #63

Open habeebp098 opened 5 months ago

habeebp098 commented 5 months ago

I have been experimenting with the alpha-beta-crown tool and encountered the following error while running it for some input data.

alpha-beta-crown) h@pillar1:~/project3/project3_2$ python /home/h/project2/alpha-beta-CROWN-main/complete_verifier/abcrown.py --onnx_path net.onnx --vnnlib_path prop_y0.vnnlb --device cpu --results_file 'abc_out_y0.txt' --no_incomplete Configurations:

general: device: cpu seed: 100 conv_mode: patches deterministic: false double_fp: false loss_reduction_func: sum record_bounds: false sparse_alpha: true save_adv_example: false precompile_jit: false complete_verifier: bab enable_incomplete_verification: false csv_name: null results_file: abc_out_y0.txt root_path: '' model: name: null path: null onnx_path: net.onnx onnx_path_prefix: '' cache_onnx_conversion: false onnx_quirks: null input_shape: null onnx_loader: default_onnx_and_vnnlib_loader onnx_optimization_flags: none data: start: 0 end: 10000 select_instance: null num_outputs: 10 mean: 0.0 std: 1.0 pkl_path: null dataset: CIFAR data_filter_path: null data_idx_file: null specification: type: lp robustness_type: verified-acc norm: .inf epsilon: null vnnlib_path: prop_y0.vnnlb vnnlib_path_prefix: '' solver: batch_size: 64 min_batch_size_ratio: 0.1 use_float64_in_last_iteration: false early_stop_patience: 10 start_save_best: 0.5 bound_prop_method: alpha-crown prune_after_crown: false crown: batch_size: 1000000000 max_crown_size: 1000000000 alpha-crown: alpha: true lr_alpha: 0.1 iteration: 100 share_slopes: false no_joint_opt: false lr_decay: 0.98 full_conv_alpha: true beta-crown: lr_alpha: 0.01 lr_beta: 0.05 lr_decay: 0.98 optimizer: adam iteration: 50 beta: true beta_warmup: true enable_opt_interm_bounds: false all_node_split_LP: false forward: refine: false dynamic: false max_dim: 10000 multi_class: multi_class_method: allclass_domain label_batch_size: 32 skip_with_refined_bound: true mip: parallel_solvers: null solver_threads: 1 refine_neuron_timeout: 15 refine_neuron_time_percentage: 0.8 early_stop: true adv_warmup: true mip_solver: gurobi bab: initial_max_domains: 1 max_domains: .inf decision_thresh: 0 timeout: 360 timeout_scale: 1 override_timeout: null get_upper_bound: false dfs_percent: 0.0 pruning_in_iteration: true pruning_in_iteration_ratio: 0.2 sort_targets: false batched_domain_list: true optimized_intermediate_layers: '' interm_transfer: true cut: enabled: false bab_cut: false lp_cut: false method: null lr: 0.01 lr_decay: 1.0 iteration: 100 bab_iteration: -1 early_stop_patience: -1 lr_beta: 0.02 number_cuts: 50 topk_cuts_in_filter: 100 batch_size_primal: 100 max_num: 1000000000 patches_cut: false cplex_cuts: false cplex_cuts_wait: 0 cplex_cuts_revpickup: true cut_reference_bounds: true fix_intermediate_bounds: false branching: method: kfsb candidates: 3 reduceop: min sb_coeff_thresh: 0.001 input_split: enable: false enhanced_bound_prop_method: alpha-crown enhanced_branching_method: naive enhanced_bound_patience: 100000000.0 attack_patience: 100000000.0 adv_check: 0 sort_domain_interval: -1 attack: enabled: false beam_candidates: 8 beam_depth: 7 max_dive_fix_ratio: 0.8 min_local_free_ratio: 0.2 mip_start_iteration: 5 mip_timeout: 30.0 adv_pool_threshold: null refined_mip_attacker: false refined_batch_size: null attack: pgd_order: before pgd_steps: 100 pgd_restarts: 30 pgd_early_stop: true pgd_lr_decay: 0.99 pgd_alpha: auto pgd_loss_mode: null enable_mip_attack: false cex_path: ./test_cex.txt attack_mode: PGD gama_lambda: 10.0 gama_decay: 0.9 check_clean: false input_split: pgd_steps: 100 pgd_restarts: 30 pgd_alpha: auto input_split_enhanced: pgd_steps: 200 pgd_restarts: 5000000 pgd_alpha: auto input_split_check_adv: pgd_steps: 5 pgd_restarts: 5 pgd_alpha: auto debug: lp_test: null

Experiments at Thu Mar 14 13:23:25 2024 on pillar1 Internal results will be saved to abc_out_y0.txt.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% idx: 0, vnnlib ID: 0 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Using onnx net.onnx Using vnnlib prop_y0.vnnlb Precompiled vnnlib file found at prop_y0.vnnlb.compiled Loading onnx net.onnx wih quirks {} /home/h/anaconda3/envs/alpha-beta-crown/lib/python3.7/site-packages/onnx2pytorch/convert/layer.py:30: UserWarning: The given NumPy array is not writable, and PyTorch does not support non-writable tensors. This means writing to this tensor will result in undefined behavior. You may want to copy the array to protect its data or make it writable before converting it to a tensor. This type of warning will be suppressed for the rest of this program. (Triggered internally at /opt/conda/conda-bld/pytorch_1646755953518/work/torch/csrc/utils/tensor_numpy.cpp:178.) layer.weight.data = torch.from_numpy(numpy_helper.to_array(weight)) /home/h/anaconda3/envs/alpha-beta-crown/lib/python3.7/site-packages/onnx2pytorch/convert/model.py:154: UserWarning: Using experimental implementation that allows 'batch_size > 1'.Batchnorm layers could potentially produce false outputs. "Using experimental implementation that allows 'batch_size > 1'." /home/h/anaconda3/envs/alpha-beta-crown/lib/python3.7/site-packages/torch/nn/functional.py:749: UserWarning: Note that order of the arguments: ceil_mode and return_indices will changeto match the args list in nn.MaxPool2d in a future release. warnings.warn("Note that order of the arguments: ceil_mode and return_indices will change" Attack parameters: initialization=uniform, steps=100, restarts=30, alpha=0.07362499833106995, initialization=uniform, GAMA=False Model output of first 5 examples: tensor([[-5.98954344, 2.28598309, 3.65411043]]) Adv example prediction (first 2 examples and 2 restarts): tensor([[[-2.10495210, -0.81939900, 3.52598715]]]) PGD attack margin (first 2 examles and 10 specs): tensor([[[1.28555310, 5.63093948]]]) number of violation: 0 Attack finished in 1.2851 seconds. PGD attack failed Total VNNLIB file length: 1, max property batch size: 1, total number of batches: 1

Properties batch 0, size 1 Remaining timeout: 358.6844446659088

Instance 0 first 10 spec matrices: [[[-1. 1. 0.]

[-1. 0. 1.]]] thresholds: [0. 0.] ###### /home/h/anaconda3/envs/alpha-beta-crown/lib/python3.7/site-packages/onnx2pytorch/operations/reshape.py:36: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs! if (shape[0] == 1 and (len(shape) == 4 or len(shape) == 2) /home/h/anaconda3/envs/alpha-beta-crown/lib/python3.7/site-packages/onnx2pytorch/operations/reshape.py:55: TracerWarning: Iterating over a tensor might cause the trace to be incorrect. Passing a tensor of different shape won't change the number of iterations executed (and might lead to errors or silently give incorrect results). shape = [x if x != 0 else input.size(i) for i, x in enumerate(shape)] Model prediction is: tensor([-5.98954344, 2.28598332, 3.65411091]) Traceback (most recent call last): File "/home/h/project2/alpha-beta-CROWN-main/complete_verifier/abcrown.py", line 650, in main() File "/home/h/project2/alpha-beta-CROWN-main/complete_verifier/abcrown.py", line 573, in main refined_betas=refined_betas, attack_images=all_adv_candidates, attack_margins=attack_margins) File "/home/h/project2/alpha-beta-CROWN-main/complete_verifier/abcrown.py", line 404, in complete_verifier rhs=rhs, timeout=timeout, attack_images=this_spec_attack_images) File "/home/h/project2/alpha-beta-CROWN-main/complete_verifier/abcrown.py", line 209, in bab timeout=timeout, refined_betas=refined_betas, rhs=rhs) File "/home/h/project2/alpha-beta-CROWN-main/complete_verifier/batch_branch_and_bound.py", line 399, in relu_bab_parallel domain, x, stop_criterion_func=stop_criterion(decision_thresh)) File "/home/h/project2/alpha-beta-CROWN-main/complete_verifier/beta_CROWN_solver.py", line 1069, in build_the_model (self.x,), share_slopes=share_slopes, c=self.c, bound_upper=False) File "/home/h/project2/alpha-beta-CROWN-main/complete_verifier/auto_LiRPA/optimized_bounds.py", line 1023, in init_slope intermediate_layer_bounds=intermediate_layer_bounds) File "/home/h/project2/alpha-beta-CROWN-main/complete_verifier/auto_LiRPA/bound_general.py", line 1321, in compute_bounds self.check_prior_bounds(final) File "/home/h/project2/alpha-beta-CROWN-main/complete_verifier/auto_LiRPA/bound_general.py", line 849, in check_prior_bounds self.check_prior_bounds(n) File "/home/h/project2/alpha-beta-CROWN-main/complete_verifier/auto_LiRPA/bound_general.py", line 849, in check_prior_bounds self.check_prior_bounds(n) File "/home/h/project2/alpha-beta-CROWN-main/complete_verifier/auto_LiRPA/bound_general.py", line 849, in check_prior_bounds self.check_prior_bounds(n) [Previous line repeated 9 more times] File "/home/h/project2/alpha-beta-CROWN-main/complete_verifier/auto_LiRPA/bound_general.py", line 857, in check_prior_bounds node.inputs[i], prior_checked=True) File "/home/h/project2/alpha-beta-CROWN-main/complete_verifier/auto_LiRPA/bound_general.py", line 959, in compute_intermediate_bounds unstable_size=unstable_size) File "/home/h/project2/alpha-beta-CROWN-main/complete_verifier/auto_LiRPA/backward_bound.py", line 156, in backward_general A, lower_b, upper_b = l.bound_backward(l.lA, l.uA, *l.inputs) File "/home/h/project2/alpha-beta-CROWN-main/complete_verifier/auto_LiRPA/operators/shape.py", line 507, in bound_backward return [(_bound_oneside(last_lA), _bound_oneside(last_uA))], 0, 0 File "/home/h/project2/alpha-beta-CROWN-main/complete_verifier/auto_LiRPA/operators/shape.py", line 505, in _bound_oneside return last_A.permute(self.perm_inv_inc_one) AttributeError: 'Patches' object has no attribute 'permute'

Strangely, the tool works perfectly for some other inputs. I have attached both the neural network and VNNLB files.
net.txt prop_y0.txt

habeebp098 commented 5 months ago

The latest version of the code is also encountering the same error for the network and VNNLIB file mentioned above.

alpha-beta-CROWN/complete_verifier$ python abcrown.py --config exp_configs/tutorial_examples/pytorch_model_with_one_vnnlib.yaml --device cpu
Configurations:

general:
  device: cpu
  seed: 100
  conv_mode: patches
  deterministic: false
  double_fp: false
  loss_reduction_func: sum
  sparse_alpha: true
  sparse_interm: true
  save_adv_example: false
  eval_adv_example: false
  show_adv_example: false
  precompile_jit: false
  complete_verifier: bab
  enable_incomplete_verification: true
  csv_name: null
  results_file: out.txt
  root_path: ''
  deterministic_opt: false
  graph_optimizer: 'Customized("custom_graph_optimizer", "default_optimizer")'
  buffer_has_batchdim: false
  save_output: false
  output_file: out.pkl
  return_optimized_model: false
model:
  name: null
  path: null
  onnx_path: net1.onnx
  onnx_path_prefix: ''
  cache_onnx_conversion: false
  debug_onnx: false
  onnx_quirks: null
  input_shape: null
  onnx_loader: default_onnx_and_vnnlib_loader
  onnx_optimization_flags: none
  onnx_vnnlib_joint_optimization_flags: none
  check_optmized: false
  flatten_final_output: false
  optimize_graph: null
  with_jacobian: false
data:
  start: 0
  end: 10000
  select_instance: null
  num_outputs: 10
  mean: 0.0
  std: 1.0
  pkl_path: null
  dataset: null
  data_filter_path: null
  data_idx_file: null
specification:
  type: lp
  robustness_type: verified-acc
  norm: .inf
  epsilon: null
  epsilon_min: 0.0
  vnnlib_path: prop_y0.vnnlb
  vnnlib_path_prefix: ''
  rhs_offset: null
solver:
  batch_size: 64
  auto_enlarge_batch_size: false
  min_batch_size_ratio: 0.1
  use_float64_in_last_iteration: false
  early_stop_patience: 10
  start_save_best: 0.5
  bound_prop_method: alpha-crown
  init_bound_prop_method: same
  prune_after_crown: false
  optimize_disjuncts_separately: false
  crown:
    batch_size: 1000000000
    max_crown_size: 1000000000
    relu_option: adaptive
  alpha-crown:
    alpha: true
    lr_alpha: 0.1
    iteration: 100
    share_alphas: false
    lr_decay: 0.98
    full_conv_alpha: true
    max_coeff_mul: .inf
    matmul_share_alphas: false
    disable_optimization: []
  invprop:
    apply_output_constraints_to: []
    tighten_input_bounds: false
    best_of_oc_and_no_oc: false
    directly_optimize: []
    oc_lr: 0.1
    share_gammas: false
  beta-crown:
    lr_alpha: 0.01
    lr_beta: 0.05
    lr_decay: 0.98
    optimizer: adam
    iteration: 50
    beta: true
    beta_warmup: true
    enable_opt_interm_bounds: false
    all_node_split_LP: false
  forward:
    refine: false
    dynamic: false
    max_dim: 10000
    reset_threshold: 1.0
  multi_class:
    label_batch_size: 32
    skip_with_refined_bound: true
  mip:
    parallel_solvers: null
    solver_threads: 1
    refine_neuron_timeout: 15
    refine_neuron_time_percentage: 0.8
    early_stop: true
    adv_warmup: true
    mip_solver: gurobi
    skip_unsafe: false
bab:
  initial_max_domains: 1
  max_domains: .inf
  decision_thresh: 0
  timeout: 360
  timeout_scale: 1
  max_iterations: -1
  override_timeout: null
  get_upper_bound: false
  pruning_in_iteration: true
  pruning_in_iteration_ratio: 0.2
  sort_targets: false
  batched_domain_list: true
  optimized_interm: ''
  interm_transfer: true
  recompute_interm: false
  sort_domain_interval: -1
  vanilla_crown: false
  cut:
    enabled: false
    implication: false
    bab_cut: false
    lp_cut: false
    method: null
    lr: 0.01
    lr_decay: 1.0
    iteration: 100
    bab_iteration: -1
    early_stop_patience: -1
    lr_beta: 0.02
    number_cuts: 50
    topk_cuts_in_filter: 1000
    batch_size_primal: 100
    max_num: 1000000000
    patches_cut: false
    cplex_cuts: false
    cplex_cuts_wait: 0
    cplex_cuts_revpickup: true
    cut_reference_bounds: true
    fix_intermediate_bounds: false
  branching:
    method: kfsb
    candidates: 3
    reduceop: min
    enable_intermediate_bound_opt: false
    branching_input_and_activation: false
    branching_input_and_activation_order: [input, relu]
    branching_input_iterations: 30
    branching_relu_iterations: 50
    nonlinear_split:
      method: shortcut
      branching_point_method: uniform
      num_branches: 2
      filter: false
      filter_beta: false
      filter_batch_size: 10000
      filter_iterations: 25
      use_min: false
      loose_tanh_threshold: null
      dynamic_bbps: false
      dynamic_options: [uniform, three_left, three_right]
    input_split:
      enable: false
      enhanced_bound_prop_method: alpha-crown
      enhanced_branching_method: naive
      enhanced_bound_patience: 100000000.0
      attack_patience: 100000000.0
      adv_check: 0
      split_partitions: 2
      sb_margin_weight: 1.0
      sb_sum: false
      bf_backup_thresh: -1
      bf_rhs_offset: 0
      bf_iters: 1000000000.0
      bf_batch_size: 100000
      bf_zero_crossing_score: false
      touch_zero_score: 0
      ibp_enhancement: false
      catch_assertion: false
      compare_with_old_bounds: false
      update_rhs_with_attack: false
      sb_coeff_thresh: 0.001
      sort_index: null
      sort_descending: true
      show_progress: false
  attack:
    enabled: false
    beam_candidates: 8
    beam_depth: 7
    max_dive_fix_ratio: 0.8
    min_local_free_ratio: 0.2
    mip_start_iteration: 5
    mip_timeout: 30.0
    adv_pool_threshold: null
    refined_mip_attacker: false
    refined_batch_size: null
attack:
  pgd_order: before
  pgd_steps: 100
  pgd_restarts: 30
  pgd_batch_size: 100000000
  pgd_early_stop: true
  pgd_lr_decay: 0.99
  pgd_alpha: auto
  pgd_alpha_scale: false
  pgd_loss_mode: null
  enable_mip_attack: false
  adv_saver: default_adv_saver
  early_stop_condition: default_early_stop_condition
  adv_example_finalizer: default_adv_example_finalizer
  pgd_loss: default_pgd_loss
  cex_path: ./test_cex.txt
  attack_mode: PGD
  attack_tolerance: 0.0
  attack_func: attack_with_general_specs
  gama_lambda: 10.0
  gama_decay: 0.9
  check_clean: false
  input_split:
    pgd_steps: 100
    pgd_restarts: 30
    pgd_alpha: auto
  input_split_enhanced:
    pgd_steps: 200
    pgd_restarts: 500000
    pgd_alpha: auto
  input_split_check_adv:
    pgd_steps: 5
    pgd_restarts: 5
    pgd_alpha: auto
    max_num_domains: 10
debug:
  view_model: false
  lp_test: null
  rescale_vnnlib_ptb: null
  test_optimized_bounds: false
  test_optimized_bounds_after_n_iterations: 0
  print_verbose_decisions: false

Experiments at Thu Jun  6 04:42:56 2024 on pillar1
Internal results will be saved to out.txt.

 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% idx: 0, vnnlib ID: 0 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Using onnx net1.onnx
Using vnnlib prop_y0.vnnlb
Precompiled vnnlib file found at prop_y0.vnnlb.compiled
Loading onnx net1.onnx wih quirks {}
/home/h/anaconda3/envs/alpha-beta-crown3/lib/python3.11/site-packages/onnx2pytorch/convert/layer.py:29: UserWarning: The given NumPy array is not writable, and PyTorch does not support non-writable tensors. This means writing to this tensor will result in undefined behavior. You may want to copy the array to protect its data or make it writable before converting it to a tensor. This type of warning will be suppressed for the rest of this program. (Triggered internally at /opt/conda/conda-bld/pytorch_1708025845868/work/torch/csrc/utils/tensor_numpy.cpp:206.)
  layer.weight.data = torch.from_numpy(numpy_helper.to_array(weight))
/home/h/anaconda3/envs/alpha-beta-crown3/lib/python3.11/site-packages/onnx2pytorch/convert/model.py:151: UserWarning: Using experimental implementation that allows 'batch_size > 1'.Batchnorm layers could potentially produce false outputs.
  warnings.warn(
Attack parameters: initialization=uniform, steps=100, restarts=30, alpha=0.07449999451637268, initialization=uniform, GAMA=False
Model output of first 5 examples:
 tensor([[-9.41845608, 11.91907406, -2.72109032]])
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:01<00:00,  1.09s/it]
Adv example prediction (first 2 examples and 2 restarts):
 tensor([[[-7.09657097,  9.00977802, -2.11059594]]])
PGD attack margin (first 2 examles and 10 specs):
 tensor([[[16.10634995,  4.98597527]]])
number of violation:  0
Attack finished in 1.0931 seconds.
PGD attack failed
/home/h/anaconda3/envs/alpha-beta-crown3/lib/python3.11/site-packages/onnx2pytorch/operations/reshape.py:36: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  if shape[0] == 1 and len(shape) in [2, 3, 4, 5] and self.quirks.get("fix_batch_size") is True:
/home/h/anaconda3/envs/alpha-beta-crown3/lib/python3.11/site-packages/onnx2pytorch/operations/reshape.py:54: TracerWarning: torch.tensor results are registered as constants in the trace. You can safely ignore this warning if you use this function to create tensors out of constant variables that would be the same every time you call this function. In any other case, this might cause the trace to be incorrect.
  if (torch.prod(torch.tensor(input.shape)) != torch.prod(shape) and len(input.size()) == len(shape) + 1
/home/h/anaconda3/envs/alpha-beta-crown3/lib/python3.11/site-packages/onnx2pytorch/operations/reshape.py:54: TracerWarning: Using len to get tensor shape might cause the trace to be incorrect. Recommended usage would be tensor.shape[0]. Passing a tensor of different shape might lead to errors or silently give incorrect results.
  if (torch.prod(torch.tensor(input.shape)) != torch.prod(shape) and len(input.size()) == len(shape) + 1
/home/h/anaconda3/envs/alpha-beta-crown3/lib/python3.11/site-packages/onnx2pytorch/operations/reshape.py:58: TracerWarning: Iterating over a tensor might cause the trace to be incorrect. Passing a tensor of different shape won't change the number of iterations executed (and might lead to errors or silently give incorrect results).
  shape = [x if x != 0 else input.size(i) for i, x in enumerate(shape)]
/home/h/project2/alpha-beta-CROWN/complete_verifier/auto_LiRPA/operators/leaf.py:192: UserWarning: The "has_batchdim" option for BoundBuffers is deprecated. It may be removed from the next release.
  warnings.warn('The "has_batchdim" option for BoundBuffers is deprecated.'
Model: BoundedModule(
  (/0): BoundInput(name=/0, inputs=[], perturbed=True)
  (/shape): BoundBuffers(name=/shape, inputs=[], perturbed=False)
  (/12): BoundParams(name=/12, inputs=[], perturbed=False)
  (/13): BoundParams(name=/13, inputs=[], perturbed=False)
  (/14): BoundParams(name=/14, inputs=[], perturbed=False)
  (/15): BoundParams(name=/15, inputs=[], perturbed=False)
  (/16): BoundParams(name=/16, inputs=[], perturbed=False)
  (/17): BoundParams(name=/17, inputs=[], perturbed=False)
  (/18): BoundParams(name=/18, inputs=[], perturbed=False)
  (/19): BoundParams(name=/19, inputs=[], perturbed=False)
  (/20): BoundParams(name=/20, inputs=[], perturbed=False)
  (/21): BoundParams(name=/21, inputs=[], perturbed=False)
  (/input): BoundTranspose(name=/input, inputs=[/0], perturbed=True)
  (/input.3): BoundConv(name=/input.3, inputs=[/input, /12, /13], perturbed=True)
  (/24): BoundRelu(name=/24, inputs=[/input.3], perturbed=True)
  (/input.7): BoundMaxPool(name=/input.7, inputs=[/24], perturbed=True)
  (/input.11): BoundConv(name=/input.11, inputs=[/input.7, /14, /15], perturbed=True)
  (/27): BoundRelu(name=/27, inputs=[/input.11], perturbed=True)
  (/input.15): BoundMaxPool(name=/input.15, inputs=[/27], perturbed=True)
  (/input.19): BoundConv(name=/input.19, inputs=[/input.15, /16, /17], perturbed=True)
  (/30): BoundRelu(name=/30, inputs=[/input.19], perturbed=True)
  (/31): BoundMaxPool(name=/31, inputs=[/30], perturbed=True)
  (/32): BoundTranspose(name=/32, inputs=[/31], perturbed=True)
  (/33): BoundConstant(name=/33, inputs=[], perturbed=False)
  (/34): BoundSplit(name=/34, inputs=[/shape, /33], perturbed=False)
  (/35): BoundSplit(name=/35, inputs=[/shape, /33], perturbed=False)
  (/36): BoundConstant(name=/36, value=tensor([0]))
  (/37): BoundSqueeze(name=/37, inputs=[/34, /36], perturbed=False)
  (/38): BoundConstant(name=/38, value=tensor([0]))
  (/39): BoundSqueeze(name=/39, inputs=[/35, /38], perturbed=False)
  (/40): BoundConstant(name=/40, value=tensor([0]))
  (/41): BoundUnsqueeze(name=/41, inputs=[/37, /40], perturbed=False)
  (/42): BoundConstant(name=/42, value=tensor([0]))
  (/43): BoundUnsqueeze(name=/43, inputs=[/39, /42], perturbed=False)
  (/44): BoundConcat(name=/44, inputs=[/41, /43], perturbed=False)
  (/45): BoundReshape(name=/45, inputs=[/32, /44], perturbed=True)
  (/input.23): BoundLinear(name=/input.23, inputs=[/45, /18, /19], perturbed=True)
  (/47): BoundRelu(name=/47, inputs=[/input.23], perturbed=True)
  (/48): BoundLinear(name=/48, inputs=[/47, /20, /21], perturbed=True)
)
Original output: tensor([[-9.41845608, 11.91907501, -2.72108960]])
Split layers:
  BoundConv(name=/input.3, inputs=[/input, /12, /13], perturbed=True): [(BoundRelu(name=/24, inputs=[/input.3], perturbed=True), 0)]
  BoundConv(name=/input.11, inputs=[/input.7, /14, /15], perturbed=True): [(BoundRelu(name=/27, inputs=[/input.11], perturbed=True), 0)]
  BoundConv(name=/input.19, inputs=[/input.15, /16, /17], perturbed=True): [(BoundRelu(name=/30, inputs=[/input.19], perturbed=True), 0)]
  BoundLinear(name=/input.23, inputs=[/45, /18, /19], perturbed=True): [(BoundRelu(name=/47, inputs=[/input.23], perturbed=True), 0)]
Nonlinear functions:
   BoundRelu(name=/24, inputs=[/input.3], perturbed=True)
   BoundMaxPool(name=/input.7, inputs=[/24], perturbed=True)
   BoundRelu(name=/27, inputs=[/input.11], perturbed=True)
   BoundMaxPool(name=/input.15, inputs=[/27], perturbed=True)
   BoundRelu(name=/30, inputs=[/input.19], perturbed=True)
   BoundMaxPool(name=/31, inputs=[/30], perturbed=True)
   BoundRelu(name=/47, inputs=[/input.23], perturbed=True)
Traceback (most recent call last):
  File "/home/h/project2/alpha-beta-CROWN/complete_verifier/abcrown.py", line 706, in <module>
    abcrown.main()
  File "/home/h/project2/alpha-beta-CROWN/complete_verifier/abcrown.py", line 638, in main
    incomplete_verification_output = self.incomplete_verifier(
                                     ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/h/project2/alpha-beta-CROWN/complete_verifier/abcrown.py", line 159, in incomplete_verifier
    global_lb, ret = model.build(
                     ^^^^^^^^^^^^
  File "/home/h/project2/alpha-beta-CROWN/complete_verifier/beta_CROWN_solver.py", line 454, in build
    lb, ub, aux_reference_bounds = self.net.init_alpha(
                                   ^^^^^^^^^^^^^^^^^^^^
  File "/home/h/project2/alpha-beta-CROWN/complete_verifier/auto_LiRPA/optimized_bounds.py", line 845, in init_alpha
    l, u = self.compute_bounds(
           ^^^^^^^^^^^^^^^^^^^^
  File "/home/h/project2/alpha-beta-CROWN/complete_verifier/auto_LiRPA/bound_general.py", line 1316, in compute_bounds
    return self._compute_bounds_main(C=C,
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/h/project2/alpha-beta-CROWN/complete_verifier/auto_LiRPA/bound_general.py", line 1414, in _compute_bounds_main
    self.check_prior_bounds(final)
  File "/home/h/project2/alpha-beta-CROWN/complete_verifier/auto_LiRPA/bound_general.py", line 879, in check_prior_bounds
    self.check_prior_bounds(n)
  File "/home/h/project2/alpha-beta-CROWN/complete_verifier/auto_LiRPA/bound_general.py", line 879, in check_prior_bounds
    self.check_prior_bounds(n)
  File "/home/h/project2/alpha-beta-CROWN/complete_verifier/auto_LiRPA/bound_general.py", line 879, in check_prior_bounds
    self.check_prior_bounds(n)
  [Previous line repeated 9 more times]
  File "/home/h/project2/alpha-beta-CROWN/complete_verifier/auto_LiRPA/bound_general.py", line 897, in check_prior_bounds
    self.compute_intermediate_bounds(
  File "/home/h/project2/alpha-beta-CROWN/complete_verifier/auto_LiRPA/bound_general.py", line 981, in compute_intermediate_bounds
    node.lower, node.upper = self.backward_general(
                             ^^^^^^^^^^^^^^^^^^^^^^
  File "/home/h/project2/alpha-beta-CROWN/complete_verifier/auto_LiRPA/backward_bound.py", line 337, in backward_general
    A, lower_b, upper_b = l.bound_backward(
                          ^^^^^^^^^^^^^^^^^
  File "/home/h/project2/alpha-beta-CROWN/complete_verifier/auto_LiRPA/operators/reshape.py", line 288, in bound_backward
    return [(_bound_oneside(last_lA), _bound_oneside(last_uA))], 0, 0
             ^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/h/project2/alpha-beta-CROWN/complete_verifier/auto_LiRPA/operators/reshape.py", line 286, in _bound_oneside
    return last_A.permute(self.perm_inv_inc_one)
           ^^^^^^^^^^^^^^
AttributeError: 'Patches' object has no attribute 'permute'