AttributeError: 'Leaky' object has no attribute 'mem'

[morenzoe]

Hi, I got this error while running benchmark.run() in the Primate_reaching_tutorial.ipynb in Google Colab. I also produced this error while running benchmark.run() in Kaggle notebook that was running fine before. I am now using neurobench ver.1.0.4 and snntorch ver.0.7.0. The error didn't happen with neurobench ver.1.0.3 and snntorch ver.0.7.0. Would be thankful for any assistance!

[ben9809]

Could you please copy and paste the full error message or upload a screenshot?

[morenzoe]

Hi, thank you for your reply. I attached a screenshot below and this is the full error message:

Running benchmark

  0%|          | 0/1 [00:08<?, ?it/s]

---------------------------------------------------------------------------

AttributeError                            Traceback (most recent call last)

[<ipython-input-11-a24b7539a050>](https://localhost:8080/#) in <cell line: 6>()
      4 # Benchmark expects the following:
      5 benchmark = Benchmark(model, test_set_loader, [], [], [static_metrics, workload_metrics])
----> 6 results = benchmark.run()
      7 print(results)

9 frames

[/usr/local/lib/python3.10/dist-packages/neurobench/benchmarks/benchmark.py](https://localhost:8080/#) in run(self, quiet, verbose, dataloader, preprocessors, postprocessors, device)
    115 
    116                 # Run model on test data
--> 117                 preds = self.model(data[0])
    118 
    119                 for alg in postprocessors:

[/usr/local/lib/python3.10/dist-packages/neurobench/models/torch_model.py](https://localhost:8080/#) in __call__(self, batch)
     31 
     32         """
---> 33         return self.net(batch)
     34 
     35     def __net__(self):

[/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py](https://localhost:8080/#) in _wrapped_call_impl(self, *args, **kwargs)
   1530             return self._compiled_call_impl(*args, **kwargs)  # type: ignore[misc]
   1531         else:
-> 1532             return self._call_impl(*args, **kwargs)
   1533 
   1534     def _call_impl(self, *args, **kwargs):

[/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py](https://localhost:8080/#) in _call_impl(self, *args, **kwargs)
   1539                 or _global_backward_pre_hooks or _global_backward_hooks
   1540                 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1541             return forward_call(*args, **kwargs)
   1542 
   1543         try:

[.../SNN2.py](https://localhost:8080/#) in forward(self, x)
     49 
     50         for sample in range(x.shape[0]):
---> 51             predictions.append(self.single_forward(x[sample, ...]))
     52 
     53         predictions = torch.stack(predictions)

[.../SNN2.py](https://localhost:8080/#) in single_forward(self, x)
     37         x = x.squeeze() # convert shape (1, input_dim) to (input_dim)
     38         cur1 = self.dropout(self.fc1(x))
---> 39         spk1, self.mem1 = self.lif1(cur1, self.mem1)
     40 
     41         cur2 = self.fc_out(spk1)

[/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py](https://localhost:8080/#) in _wrapped_call_impl(self, *args, **kwargs)
   1530             return self._compiled_call_impl(*args, **kwargs)  # type: ignore[misc]
   1531         else:
-> 1532             return self._call_impl(*args, **kwargs)
   1533 
   1534     def _call_impl(self, *args, **kwargs):

[/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py](https://localhost:8080/#) in _call_impl(self, *args, **kwargs)
   1569                                 )
   1570                     else:
-> 1571                         args_result = hook(self, args)
   1572                         if args_result is not None:
   1573                             if not isinstance(args_result, tuple):

[/usr/local/lib/python3.10/dist-packages/neurobench/benchmarks/hooks.py](https://localhost:8080/#) in pre_hook_fn(self, layer, input)
     50         self.activation_inputs.append(input)
     51         if self.spiking:
---> 52             self.pre_fire_mem_potential.append(layer.mem)
     53 
     54     def hook_fn(self, layer, input, output):

[/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py](https://localhost:8080/#) in __getattr__(self, name)
   1707             if name in modules:
   1708                 return modules[name]
-> 1709         raise AttributeError(f"'{type(self).__name__}' object has no attribute '{name}'")
   1710 
   1711     def __setattr__(self, name: str, value: Union[Tensor, 'Module']) -> None:

AttributeError: 'Leaky' object has no attribute 'mem'

[ben9809]

Hi @morenzoe, thank you for bringing this to our attention. I managed to solve the issue, once we review and merge this pull request it should be fixed!

[morenzoe]

Thank you so much! Looking forward for the update 😄

[jasonlyik]

Bug fixed in latest version 1.0.5