Haojin Yang

Extension of flexibility in federated learning scenarios

3

**Background Information**: My team and I (from Hasso Plattner Institute) will make a new MOOC lecture on the https://ki-campus.org/ platform. I want to use Sedna to design two practical tasks...

binary conv arm中input bit-packing像素顺序问题

6

Thanks for the great work！ 在看了代码后，我发现bolt的实现中可能存在一些问题，但也不是很确定，想请教一下： 比如这里：[convolution_xnor_A55](https://github.com/huawei-noah/bolt/blob/c3eb7a22e4f7acc2cc450606d1875666d4b11574/compute/tensor/src/cpu/arm/bnn/convolution_xnor_A55.cpp#L69) 中 [transformFromHalf](https://github.com/huawei-noah/bolt/blob/c3eb7a22e4f7acc2cc450606d1875666d4b11574/common/uni/include/data_type.h#L90)本质上是对input tensor做了bit-packing，但潜在的问题是它是按照row-wise的顺序的方式对整个input tensor进行像素访问。而通常实现binary 矩阵乘法时，需要先做im2col，将input变成patches，每一个patch包含和weight filter大小一样的子矩阵，然后再对每一个patch中的像素进行bit-packing。这里可以看到问题了，在bolt的实现中input tensor的bit-packing是不是做早了？packing到一起的像素并不属于卷积时的相同的patch？希望得到您的答复，谢谢！

error occurred while saving states

1

thanks a lot for your excellent implementation! I just got an issue when trying to call the _save_states()_ function of gluon trainer after the training finished. error log: ... return...