-
I found this paper yesterday about Binary Networks/Convolutions and XNOR-NET: http://arxiv.org/pdf/1603.05279v1.pdf (XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks).
Any…
-
if train_with_fp_copy flag is false, does quant_aware training still works? (In my try, the two cases's performance differs a lot)
-
Hi All,
I modify DoReFa to use VGG11 with Cifar10. I evaluate three cases W,A,G=8,8,32, W,A,G=4,4,32, W,A,G=2,2,32 with 50 epochs (BatchNorm is used for large batch size=256). The unexpected proble…
-
I want to train or optimize only certain layers(fct layer) of DoRefo-net , so I use fine-turning in alexnet-dorefa.py by following way: Using minimize method to take an optional argument var_list, a l…
-
Hi, @mohdumar644
Exciting to see the implement of dorefa with pytorch. I wonder if you have some accuracy on datasets such as cifar10/100/imagenet and networks of resnet10/20/alexnet to be shared?
…
-
Recently, I wanna to use the Tensorpack Dorefanet examples to train cifar100 dataset from scratch. However, the start initial learning rate needs to set very small such as 0.0001 even it works.
Ten…
-
I am using "DumpTensors" to retrieve tensor values duing Training. How can I get the same tensor's value during inference? As far as I know, this callbacks only works during training.
Thanks a lot.
…
-
Is there any plan to implement the stochastic quantization for gradients as described in the DoReFa-Net paper?
-
In DoreFa-Net, I train the alexnet using the command `./alexnet-dorefa.py --dorefa 1,8,32 --data /workspace/home/zhoucc/tensorflow/Dataset/ImageNet_ILSVRC2012/PATH/ --gpu 0,1
`
**Messages:**
[052…
-
first thanks for you paper.
I want to know ,how to quantization the weights and input about about the result ??
can you tell me some code about quantization ?