Issues
- 2
Speed up training
#15 opened by bigaidream - 2
Support batchSize>1
#12 opened by bigaidream - 5
Optimize CUDA data transfer
#13 opened by bigaidream - 3
Replace the forward pass with `cunn`
#8 opened by bigaidream - 0
- 0
Add weight decay in forward pass for L2 tuning
#11 opened by bigaidream - 0
Verify effectiveness on ImageNet
#10 opened by bigaidream - 0
Verify effectiveness on CIFAR-10
#9 opened by bigaidream - 3
Tuning learning rates
#6 opened by bigaidream - 0
GPU support
#5 opened by bigaidream - 0
Multiple meta-iterations support
#4 opened by bigaidream - 1
The drmad loss grows to NaN
#3 opened by nicholas-leonard - 0
hypergrads are always zeros
#1 opened by bigaidream