torchstat
torchstat copied to clipboard
Swall0w
Model analyzer in PyTorch
```bash python setup.py build ``` outputs: ```text /usr/lib/python3.10/distutils/dist.py:274: UserWarning: Unknown distribution option: 'tests_requires' warnings.warn(msg) running build running build_py creating build creating build/lib creating build/lib/torchstat copying torchstat/statistics.py -> build/lib/torchstat copying torchstat/stat_tree.py...
Hello! Thanks for your great work at first. I want to check my Spiking Neural Network model's GPU using `torchstat`, but it seems not to support inputs with SNN layers......
In the total result row, the results of total memory_wirte and total memory_read are incorrect, which just show the last layer's result. https://github.com/Swall0w/torchstat/blob/b52a3b06c2c54c2d09ade1a18cf6c4ca5dc27510/torchstat/reporter.py#L57-L62 Here should be `total_mread` and 'total_mwrite' rather...
Is the calculation of memory occupation the sum of the sizes of all feature maps? Or is it the calculation of peak memory ?
Hey Swall0w, I have a question regarding the calculation of Macs and Flops. My understanding is that the number of flops should be about twice the number of Macs. Since...
I think when bias is True, Madd shouls be 72. `class StatModel(nn.Module): def __init__(self): super(StatModel, self).__init__() self.layers_bias_false = nn.Linear(12, 3, bias=False) self.layers_bias_true = nn.Linear(12, 3, bias=True) def forward(self, x): x...
When I run the stat() code, it gives "assert module.__class__ in self._origin_call". I print out the entire log and I find out this module has 3 children and still being...
if I want to get the calculation analysis of a fc layer, how can I use this tool? import torch.nn as nn import torch from torchstat import stat class Net(nn.Module):...
