XNOR-Net-PyTorch icon indicating copy to clipboard operation
XNOR-Net-PyTorch copied to clipboard

PyTorch Implementation of XNOR-Net

Results 24 XNOR-Net-PyTorch issues
Sort by recently updated
recently updated
newest added

to fix "Legacy autograd function with non-static forward method is deprecated. " RuntimeError: Legacy autograd function with non-static forward method is deprecated. Please use new-style autograd function with static forward...

Has the research accelerated in the actual deployment of xnor-net been realized?

首先,非常感谢您开源了您的XNOR-pytorch代码。其次,我注意到您在更新单精度权重时,对于权重的梯度乘了一些系数: `self.target_modules[index].grad.data = m.add(m_add).mul(1.0-1.0/s[1]).mul(n)` `self.target_modules[index].grad.data = self.target_modules[index].grad.data.mul(1e+9)` 关于这些系数,我没有在原文中找到相应的描述,想问一下您为什么对梯度进行了这样的变换。

这是在torch1.4之后的新要求,改变了torch.autograd.Function的调用方法,需要使用.apply来调用。 解决方法,在XNOR-Net-PyTorch-master\MNIST\models下的LeNet_5.py第53行改为 x = BinActive.apply(x)即可正常运行 change LeNet_5.py line53 to 'x = BinActive.apply(x)' to slove problem as following: RuntimeError: Legacy autograd function with non-static forward method is deprecated

感谢您开源了XNOR的pytorch代码 有一些疑问,您当时使用的是哪个pytorch版本呢

![image](https://user-images.githubusercontent.com/87064809/126381413-01522e20-fb7b-4f68-85aa-9605924ea66d.png) ![image](https://user-images.githubusercontent.com/87064809/126381556-c4867018-4a0d-4a85-9c1d-61e62edd33dd.png) How Differential scaling facter alpha? i think differential alpha = 1 what is error of my thought? please reply me..

Hi. Thank you for your uploading. I have downloaded NIN cifar10 pretrained model and load it. Then, I run the inference evaluation. At this time. the accuracy is just only...

Hi @jiecaoyu Thanks for the nice project. I am new to this raw level coding. I have a question, if my input data is already binarised, in which part of...

In [this commit ](https://github.com/jiecaoyu/XNOR-Net-PyTorch/commit/6347cadb053ef8e2c34cc18e1f69017ce39ce3c2#diff-d1f0739aecc8d9260607ade763e0890a8cbc9d4486d810fc59dc04dbb4a4950d) the LeNet5 Implementation was changed. The [current link](https://drive.google.com/open?id=0B-7I62GOSnZ8R3Jzd0ozdzlJUk0) to download a pre-trained LeNet5 model in README.md is outdated and generates a size mismatch when loading the...

Hi, In your gradient update, you multiplied mul(1.0-1.0/s[1]).mul(n) to the gradient, which does not appear in your note.pdf. Could you explain why you multiplied these two terms? Thanks