github.com/torch/nn.git - Unnamed repository; edit this file 'description' to name the repository.

Age	Commit message (Collapse)	Author
2017-02-03	Support index parameters with Cuda*Tensor.	Gregory Chanan
	These modules check the type explicitly for CudaTensor rather than the regular expression. There are other modules that have this identical check, but do operations on the resulting type which may or may not be support (i.e. require more checking), so I didn't change them. These should be safe because the index type is independent of the type being checking.
2016-10-02	More broken compatibility with cutorch 1.0	Martin Simonovsky
	More broken compatibility with cutorch 1.0
2016-08-17	fixing nn.Normalize for new cutorch types	soumith

2016-03-23	Reuse same state tensors on forward/backward	Francisco Massa

2016-02-09	nn.clearState	Sergey Zagoruyko

2015-12-24	Remove bmm and baddbmm from Normalize, because they allocate memory, causing ↵	Adam Lerer
	sync on CUDA.
2015-11-16	Normalize supports infinity (maximum) norm	Francisco Massa

2015-10-24	Fix possible division by zero in Normalize	Francisco Massa
	When p < 2, if the input is 0 there was division by zero. Add a small dampening factor to avoid this. Also, small optimizations for different p.
2015-10-20	Fix modification of output in nn.Normalize	Adria Puigdomenech

2015-10-20	Revert "Fix modification of output in nn.Normalize"	Soumith Chintala

2015-10-16	Fix modification of output in nn.Normalize	Adria Puigdomenech

2015-10-11	Speedup and reduce memory usage in Normalize	Francisco Massa
	It's almost 40x faster on CPU, and doesn't use extra memory. Should be faster on GPU as well
2015-09-13	Adds Normalize	Francisco Massa