登录
当前位置： 首页 > 工具软件 > SpikingJelly > 使用案例 >

spikingjelly中的ANN2SNN程序20201221是好使的

许正平
2023-12-01

spikingjelly的最新版本，我当时使用的是20210330，中间存在onnx的包调用错误，难以解决。

通过调试20201221，发现这个版本的是好使的，但是也有一些是需要调试的。
1.首先应该训练出一个ANN模型出来，此时main函数中应该不填入任何信息，即

if __name__ == '__main__':
    main()
    # main('./log-cnn_mnist1622169436.7251258')

2.当训练出ANN模型后，会自动生成一个文件夹
如

./log-cnn_mnist1622169436.7251258

这个文件夹中存有大量的ANN和SNN模型
3.

 utils.pytorch_ann2snn(model_name=model_name,
                              norm_tensor=norm_tensor,
                              test_data_loader=test_data_loader,
                              device=device,
                              T=T,
                              log_dir=log_dir,
                              config=config
                              )

这里面有好几个坑：
3.1
原来的代码中,第二个参数写的不是norm_tensor，好像是test_tensor，这个容易发现，按着ctrl 点击pytorch_ann2snn就能很快找到错误。
3.2
utils.py 里面有大量的function
其中有一个pytorch_ann2snn,这个function里面有一行是使用val_ann的
正确代码如下：

ann_acc = val_ann(net=parsed_ann, loss_function=nn.CrossEntropyLoss(),device=device, data_loader=test_data_loader)

原来的代码中，好像是少了一个loss_function的参数
3.3
还是val_ann的问题，但是我忘记了。。。。

反正最后是成功了
控制台输出代码如下：

All the temp files are saved to  ./log-cnn_mnist1622169436.7251258
ann2snn config:
	 {'simulation': {'reset_to_zero': False, 'encoder': {'possion': False}, 'avg_pool': {'has_neuron': True}, 'max_pool': {'if_spatial_avg': False, 'if_wta': False, 'momentum': None}}, 'parser': {'robust_norm': True}}
Directly load model cnn_mnist.pkl
Using 120 pictures as norm set
Load best model for Model:cnn_mnist...
ANN Validating Accuracy:0.986
Save model to: ./log-cnn_mnist1622169436.7251258\parsed_cnn_mnist.pkl
Using robust normalization...
normalize with bias...
Print Parsed ANN model Structure:
Pytorch_Parser(
  (network): Sequential(
    (0): Conv2d(1, 32, kernel_size=(3, 3), stride=(1, 1))
    (1): ReLU()
    (2): AvgPool2d(kernel_size=2, stride=2, padding=0)
    (3): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1))
    (4): ReLU()
    (5): AvgPool2d(kernel_size=2, stride=2, padding=0)
    (6): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1))
    (7): ReLU()
    (8): AvgPool2d(kernel_size=2, stride=2, padding=0)
    (9): Flatten(start_dim=1, end_dim=-1)
    (10): Linear(in_features=32, out_features=10, bias=True)
    (11): ReLU()
  )
)
Save model to: ./log-cnn_mnist1622169436.7251258\normalized_cnn_mnist.pkl
Print Simulated SNN model Structure:
PyTorch_Converter(
  (network): Sequential(
    (0): Conv2d(1, 32, kernel_size=(3, 3), stride=(1, 1))
    (1): IFNode(
      v_threshold=1.0, v_reset=None, detach_reset=False
      (surrogate_function): Sigmoid(alpha=1.0, spiking=True)
    )
    (2): AvgPool2d(kernel_size=2, stride=2, padding=0)
    (3): IFNode(
      v_threshold=1.0, v_reset=None, detach_reset=False
      (surrogate_function): Sigmoid(alpha=1.0, spiking=True)
    )
    (4): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1))
    (5): IFNode(
      v_threshold=1.0, v_reset=None, detach_reset=False
      (surrogate_function): Sigmoid(alpha=1.0, spiking=True)
    )
    (6): AvgPool2d(kernel_size=2, stride=2, padding=0)
    (7): IFNode(
      v_threshold=1.0, v_reset=None, detach_reset=False
      (surrogate_function): Sigmoid(alpha=1.0, spiking=True)
    )
    (8): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1))
    (9): IFNode(
      v_threshold=1.0, v_reset=None, detach_reset=False
      (surrogate_function): Sigmoid(alpha=1.0, spiking=True)
    )
    (10): AvgPool2d(kernel_size=2, stride=2, padding=0)
    (11): IFNode(
      v_threshold=1.0, v_reset=None, detach_reset=False
      (surrogate_function): Sigmoid(alpha=1.0, spiking=True)
    )
    (12): Flatten(start_dim=1, end_dim=-1)
    (13): Linear(in_features=32, out_features=10, bias=True)
    (14): IFNode(
      v_threshold=1.0, v_reset=None, detach_reset=False
      (surrogate_function): Sigmoid(alpha=1.0, spiking=True)
    )
  )
)
100%|██████████| 100/100 [00:00<00:00, 285.67it/s]
[SNN Simulating... 1.00%] Acc:0.990
100%|██████████| 100/100 [00:00<00:00, 374.17it/s]
[SNN Simulating... 2.00%] Acc:0.995
 78%|███████▊  | 78/100 [00:00<00:00, 383.98it/s][SNN Simulating... 3.00%] Acc:0.990
100%|██████████| 100/100 [00:00<00:00, 383.84it/s]
100%|██████████| 100/100 [00:00<00:00, 387.40it/s]
[SNN Simulating... 4.00%] Acc:0.993
100%|██████████| 100/100 [00:00<00:00, 380.71it/s]
[SNN Simulating... 5.00%] Acc:0.992
 77%|███████▋  | 77/100 [00:00<00:00, 386.11it/s][SNN Simulating... 6.00%] Acc:0.992
100%|██████████| 100/100 [00:00<00:00, 382.31it/s]
 77%|███████▋  | 77/100 [00:00<00:00, 379.92it/s][SNN Simulating... 7.00%] Acc:0.991
100%|██████████| 100/100 [00:00<00:00, 382.12it/s]
100%|██████████| 100/100 [00:00<00:00, 380.98it/s]
[SNN Simulating... 8.00%] Acc:0.990
100%|██████████| 100/100 [00:00<00:00, 383.18it/s]
[SNN Simulating... 9.00%] Acc:0.991
100%|██████████| 100/100 [00:00<00:00, 380.64it/s]
[SNN Simulating... 10.00%] Acc:0.991
 78%|███████▊  | 78/100 [00:00<00:00, 382.40it/s][SNN Simulating... 11.00%] Acc:0.989
100%|██████████| 100/100 [00:00<00:00, 382.72it/s]
100%|██████████| 100/100 [00:00<00:00, 387.78it/s]
[SNN Simulating... 12.00%] Acc:0.988
 77%|███████▋  | 77/100 [00:00<00:00, 380.84it/s][SNN Simulating... 13.00%] Acc:0.989
100%|██████████| 100/100 [00:00<00:00, 382.14it/s]
 77%|███████▋  | 77/100 [00:00<00:00, 379.64it/s][SNN Simulating... 14.00%] Acc:0.989
100%|██████████| 100/100 [00:00<00:00, 380.64it/s]
 78%|███████▊  | 78/100 [00:00<00:00, 386.31it/s][SNN Simulating... 15.00%] Acc:0.990
100%|██████████| 100/100 [00:00<00:00, 384.24it/s]
100%|██████████| 100/100 [00:00<00:00, 382.49it/s]
[SNN Simulating... 16.00%] Acc:0.989
100%|██████████| 100/100 [00:00<00:00, 383.12it/s]
[SNN Simulating... 17.00%] Acc:0.989
100%|██████████| 100/100 [00:00<00:00, 381.43it/s]
[SNN Simulating... 18.00%] Acc:0.989
100%|██████████| 100/100 [00:00<00:00, 384.29it/s]
[SNN Simulating... 19.00%] Acc:0.989
100%|██████████| 100/100 [00:00<00:00, 383.13it/s]
[SNN Simulating... 20.00%] Acc:0.989
 77%|███████▋  | 77/100 [00:00<00:00, 383.32it/s][SNN Simulating... 21.00%] Acc:0.988
100%|██████████| 100/100 [00:00<00:00, 383.44it/s]
100%|██████████| 100/100 [00:00<00:00, 384.28it/s]
[SNN Simulating... 22.00%] Acc:0.988
 78%|███████▊  | 78/100 [00:00<00:00, 382.79it/s][SNN Simulating... 23.00%] Acc:0.988
100%|██████████| 100/100 [00:00<00:00, 382.40it/s]
100%|██████████| 100/100 [00:00<00:00, 377.74it/s]
[SNN Simulating... 24.00%] Acc:0.988
 78%|███████▊  | 78/100 [00:00<00:00, 382.20it/s][SNN Simulating... 25.00%] Acc:0.987
100%|██████████| 100/100 [00:00<00:00, 383.59it/s]
100%|██████████| 100/100 [00:00<00:00, 385.62it/s]
[SNN Simulating... 26.00%] Acc:0.987
 78%|███████▊  | 78/100 [00:00<00:00, 383.72it/s][SNN Simulating... 27.00%] Acc:0.987
100%|██████████| 100/100 [00:00<00:00, 384.30it/s]
 77%|███████▋  | 77/100 [00:00<00:00, 378.39it/s][SNN Simulating... 28.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 383.11it/s]
 77%|███████▋  | 77/100 [00:00<00:00, 380.10it/s][SNN Simulating... 29.00%] Acc:0.987
100%|██████████| 100/100 [00:00<00:00, 382.31it/s]
 78%|███████▊  | 78/100 [00:00<00:00, 385.14it/s][SNN Simulating... 30.00%] Acc:0.987
100%|██████████| 100/100 [00:00<00:00, 384.02it/s]
 78%|███████▊  | 78/100 [00:00<00:00, 380.23it/s][SNN Simulating... 31.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 381.53it/s]
100%|██████████| 100/100 [00:00<00:00, 384.25it/s]
[SNN Simulating... 32.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 386.27it/s]
[SNN Simulating... 33.00%] Acc:0.986
 78%|███████▊  | 78/100 [00:00<00:00, 381.66it/s][SNN Simulating... 34.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 381.08it/s]
 78%|███████▊  | 78/100 [00:00<00:00, 386.44it/s][SNN Simulating... 35.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 383.48it/s]
100%|██████████| 100/100 [00:00<00:00, 384.00it/s]
[SNN Simulating... 36.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 385.31it/s]
[SNN Simulating... 37.00%] Acc:0.985
 78%|███████▊  | 78/100 [00:00<00:00, 383.32it/s][SNN Simulating... 38.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 382.76it/s]
100%|██████████| 100/100 [00:00<00:00, 382.12it/s]
[SNN Simulating... 39.00%] Acc:0.986
 77%|███████▋  | 77/100 [00:00<00:00, 379.25it/s][SNN Simulating... 40.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 380.26it/s]
100%|██████████| 100/100 [00:00<00:00, 381.43it/s]
[SNN Simulating... 41.00%] Acc:0.985
 78%|███████▊  | 78/100 [00:00<00:00, 385.23it/s][SNN Simulating... 42.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 385.14it/s]
100%|██████████| 100/100 [00:00<00:00, 382.23it/s]
[SNN Simulating... 43.00%] Acc:0.985
 78%|███████▊  | 78/100 [00:00<00:00, 381.57it/s][SNN Simulating... 44.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 381.63it/s]
 80%|████████  | 80/100 [00:00<00:00, 382.68it/s][SNN Simulating... 45.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 382.89it/s]
100%|██████████| 100/100 [00:00<00:00, 381.87it/s]
[SNN Simulating... 46.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 382.94it/s]
[SNN Simulating... 47.00%] Acc:0.986
 78%|███████▊  | 78/100 [00:00<00:00, 381.88it/s][SNN Simulating... 48.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 381.80it/s]
 78%|███████▊  | 78/100 [00:00<00:00, 380.89it/s][SNN Simulating... 49.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 381.39it/s]
100%|██████████| 100/100 [00:00<00:00, 385.67it/s]
[SNN Simulating... 50.00%] Acc:0.987
 77%|███████▋  | 77/100 [00:00<00:00, 380.18it/s][SNN Simulating... 51.00%] Acc:0.987
100%|██████████| 100/100 [00:00<00:00, 380.93it/s]
 78%|███████▊  | 78/100 [00:00<00:00, 384.34it/s][SNN Simulating... 52.00%] Acc:0.987
100%|██████████| 100/100 [00:00<00:00, 384.01it/s]
100%|██████████| 100/100 [00:00<00:00, 381.28it/s]
[SNN Simulating... 53.00%] Acc:0.987
 78%|███████▊  | 78/100 [00:00<00:00, 382.42it/s][SNN Simulating... 54.00%] Acc:0.987
100%|██████████| 100/100 [00:00<00:00, 383.27it/s]
100%|██████████| 100/100 [00:00<00:00, 380.66it/s]
[SNN Simulating... 55.00%] Acc:0.987
 77%|███████▋  | 77/100 [00:00<00:00, 383.32it/s][SNN Simulating... 56.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 381.38it/s]
100%|██████████| 100/100 [00:00<00:00, 384.44it/s]
[SNN Simulating... 57.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 380.93it/s]
[SNN Simulating... 58.00%] Acc:0.987
 77%|███████▋  | 77/100 [00:00<00:00, 383.83it/s][SNN Simulating... 59.00%] Acc:0.987
100%|██████████| 100/100 [00:00<00:00, 382.44it/s]
 77%|███████▋  | 77/100 [00:00<00:00, 379.82it/s][SNN Simulating... 60.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 380.56it/s]
 77%|███████▋  | 77/100 [00:00<00:00, 381.51it/s][SNN Simulating... 61.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 378.22it/s]
100%|██████████| 100/100 [00:00<00:00, 378.32it/s]
[SNN Simulating... 62.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 381.82it/s]
[SNN Simulating... 63.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 380.16it/s]
[SNN Simulating... 64.00%] Acc:0.986
 77%|███████▋  | 77/100 [00:00<00:00, 379.44it/s][SNN Simulating... 65.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 381.32it/s]
100%|██████████| 100/100 [00:00<00:00, 380.55it/s]
[SNN Simulating... 66.00%] Acc:0.986
100%|██████████| 100/100 [00:00<00:00, 377.32it/s]
[SNN Simulating... 67.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 380.96it/s]
[SNN Simulating... 68.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 380.04it/s]
[SNN Simulating... 69.00%] Acc:0.985
 78%|███████▊  | 78/100 [00:00<00:00, 380.84it/s][SNN Simulating... 70.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 381.85it/s]
100%|██████████| 100/100 [00:00<00:00, 381.54it/s]
[SNN Simulating... 71.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 383.40it/s]
[SNN Simulating... 72.00%] Acc:0.985
 78%|███████▊  | 78/100 [00:00<00:00, 382.86it/s][SNN Simulating... 73.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 382.38it/s]
 77%|███████▋  | 77/100 [00:00<00:00, 381.36it/s][SNN Simulating... 74.00%] Acc:0.984
100%|██████████| 100/100 [00:00<00:00, 382.45it/s]
 77%|███████▋  | 77/100 [00:00<00:00, 380.61it/s][SNN Simulating... 75.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 381.50it/s]
100%|██████████| 100/100 [00:00<00:00, 384.54it/s]
[SNN Simulating... 76.00%] Acc:0.985
 76%|███████▌  | 76/100 [00:00<00:00, 375.56it/s][SNN Simulating... 77.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 377.99it/s]
100%|██████████| 100/100 [00:00<00:00, 384.83it/s]
[SNN Simulating... 78.00%] Acc:0.985
 72%|███████▏  | 72/100 [00:00<00:00, 346.83it/s][SNN Simulating... 79.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 363.56it/s]
 77%|███████▋  | 77/100 [00:00<00:00, 380.04it/s][SNN Simulating... 80.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 380.84it/s]
100%|██████████| 100/100 [00:00<00:00, 382.03it/s]
[SNN Simulating... 81.00%] Acc:0.985
 77%|███████▋  | 77/100 [00:00<00:00, 379.45it/s][SNN Simulating... 82.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 380.99it/s]
100%|██████████| 100/100 [00:00<00:00, 381.97it/s]
[SNN Simulating... 83.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 382.07it/s]
[SNN Simulating... 84.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 382.60it/s]
[SNN Simulating... 85.00%] Acc:0.985
 83%|████████▎ | 83/100 [00:00<00:00, 380.87it/s][SNN Simulating... 86.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 380.21it/s]
 78%|███████▊  | 78/100 [00:00<00:00, 380.85it/s][SNN Simulating... 87.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 378.54it/s]
 76%|███████▌  | 76/100 [00:00<00:00, 373.97it/s][SNN Simulating... 88.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 375.22it/s]
100%|██████████| 100/100 [00:00<00:00, 382.67it/s]
[SNN Simulating... 89.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 384.07it/s]
[SNN Simulating... 90.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 385.50it/s]
[SNN Simulating... 91.00%] Acc:0.985
 78%|███████▊  | 78/100 [00:00<00:00, 384.77it/s][SNN Simulating... 92.00%] Acc:0.984
100%|██████████| 100/100 [00:00<00:00, 382.92it/s]
 77%|███████▋  | 77/100 [00:00<00:00, 381.75it/s][SNN Simulating... 93.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 381.70it/s]
100%|██████████| 100/100 [00:00<00:00, 381.89it/s]
[SNN Simulating... 94.00%] Acc:0.985
 78%|███████▊  | 78/100 [00:00<00:00, 382.47it/s][SNN Simulating... 95.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 381.82it/s]
100%|██████████| 100/100 [00:00<00:00, 382.68it/s]
[SNN Simulating... 96.00%] Acc:0.985
 78%|███████▊  | 78/100 [00:00<00:00, 383.37it/s][SNN Simulating... 97.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 383.90it/s]
100%|██████████| 100/100 [00:00<00:00, 381.86it/s]
[SNN Simulating... 98.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 378.68it/s]
[SNN Simulating... 99.00%] Acc:0.985
100%|██████████| 100/100 [00:00<00:00, 383.33it/s]
[SNN Simulating... 100.00%] Acc:0.985
SNN Simulating Accuracy:0.985
Summary:	ANN Accuracy:98.6100%  	SNN Accuracy:98.5000% [Decreased 0.1100%]
 类似资料：

相关阅读

PyQt应用程序中的线程：使用Qt线程还是Python线程？使用Enum的序数值对Java中的数组进行索引是否是不好的做法？在Angular中使用jQuery是不好的做法吗？Java中的守护程序线程是什么？什么是Java中的守护程序线程？

相关文章

同程面试 9.21 超好体验ASP 程序秋招首挂 但是字节 好难受使用VC6.0编写C语言程序使用CLion编写C语言程序

相关问答

在Python中停止程序并保存数据的最好方法是什么？在应用程序类上使用registerReceiver是一种良好的已知做法吗？前端 - 写好的H5如何使用小程序运行？Google drive是多租户应用程序的一个好例子吗？使用lateinit是最好的方法吗？

相关文档

Java 程序员眼中的 Linux好用的中文速查表Rexx 重构的扩展执行程序 中文教程更好的 Java 中英文对照疯狂的程序员
快捷导航： 新手教程 算法原理 架构设计 Java进阶 数据库进阶 大厂专栏 面试经验 编程笔记 编程问答 所有专题 文档资料 工具软件 电子书籍 小牛导航
在线工具： 房贷计算器 个税计算器 Linux命令查询 Json格式化 正则表达式 颜色转换 AES加解密 SHA1加密 MD5加密 毒鸡汤 字数统计 随机密码生成 进制转换 Base64编解码 励志句子
Copyright © 2019-2024 小牛知识库@xnip.cn. All Rights Reserved.