assert len(graph.nodes) == len(graph_check.nodes)

[HuYue233]

Describe the bug:
I compress my model with L1NormPruner and speedup it, then an error occurs. How can I solve this problem?

This is error:

I checked it, but I don't know how to solve this problem：

This is the code for the pruning part of my project:

device = torch.device("cpu")
inputs = torch.randn((1, 3, 768, 768))
model_path = 'weights/yolov3_cqxdq_total_300_300.pt'
pruner_model_path = 'weights/yolov3_cqxdq_pruner_weights.pth'
config_list = [{'sparsity': 0.6, 'op_types': ['Conv2d']}]

model = attempt_load(model_path, map_location=device)  # load FP32 model

from nni.compression.pytorch.pruning import L1NormPruner

pruner = L1NormPruner(model, config_list)
_, masks = pruner.compress()
for name, mask in masks.items():
    print(name, ' sparsity : ', '{:.2}'.format(mask['weight'].sum() / mask['weight'].numel()))
pruner._unwrap_model()

from nni.compression.pytorch.speedup.v2 import ModelSpeedup

m_speedup = ModelSpeedup(model, inputs, masks, device, batch_size=2)
m_speedup.speedup_model()

This is the structure and forward of my model：
https://github.com/ultralytics/yolov3
Due to the high version of Pytorch, I have made modifications to this:Originally posted by @EdwardAndersonMcDermott in ultralytics/yolov5#6948 (comment)

And I deleted the control-flow:

Environment:
NNI version: v3.0rc1
Training service (local|remote|pai|aml|etc): local
Python version: 3.8.5
PyTorch version: 1.11.0
Cpu or cuda version: cpu

Reproduce the problem

• Code|Example:

• How to reproduce:

[J-shang]

hello @HuYue233 , many reasons may cause the nodes of the two graphs to not match, some are caused by randomness, some are caused by dead nodes, we have updated the logic of cleaning up dead nodes in the latest version, you can try to install from source code and try again to see if the problem is caused by dead nodes.

[Lijiaoa]

Could you give us some updates? @HuYue233

[HuYue233]

I commented out line 47 of the code.