Age | Commit message (Collapse) | Author | |
---|---|---|---|
2015-07-01 | fixing torch.random(a,b) for a,b >= 2^31 | Clement Farabet | |
- current implementation casts random number to int | |||
2015-06-11 | Merge pull request #252 from d11/torch_range | Soumith Chintala | |
Tweak torch.range to be more numerically robust. | |||
2015-06-03 | Tweak torch.range to be more numerically robust. | Dan Horgan | |
Otherwise it can produce mildly surprising behaviour, in the floating-point setting, due to rounding errors. | |||
2015-05-29 | Add support for Lua 5.2 | Sam Gross | |
Includes aliases to functions deprecated in Lua 5.2, such as math.log10 and unpack, and loadstring | |||
2015-05-19 | Add scatter and gather operations. | Dominik Grewe | |
Functions to read to / write from tensors, selecting elements from each slice along the specified dimension given an index tensor. | |||
2015-05-12 | Add LAPACK QR decomposition to Torch. | Dan Horgan | |
Three functions are added: torch.geqrf() and torch.orgqr(), which call the corresponding LAPACK functions directly, and torch.qr(), which is just a convenience function that calls them both correctly and puts the results in a more intuitive form. | |||
2015-03-23 | Added kthvalue and median operations, tests, doc | Georg Ostrovski | |
2015-02-04 | Make bmm use baddbmm. | Dominik Grewe | |
Just like mm uses addmm, we can write bmm in terms of baddbmm. This way the two behave the same in terms of when the output is resized. Also gets rid of unnecessary zeroing out of output of mm and mv. | |||
2015-01-22 | Renamed baddmm -> addbmm; added baddbmm. | Clement Farabet | |
- addbmm is a batch MM + reduce add - baddbmm is a batch MM + batch add | |||
2015-01-21 | Add batch matrix-matrix multiplication with accumulation. | Clement Farabet | |
- behavior is similar to torch.addmm, except that both input matrices have an additional dimension, as in torch.bmm - tests cover all use cases | |||
2015-01-21 | Added alternative version of pow function | Georg Ostrovski | |
2015-01-21 | Added element-wise power fct cpow | Georg Ostrovski | |
2015-01-14 | Add batch matrix-matrix multiplication. | Dominik Grewe | |
Add torch.bmm, which expects two 3D tensors as inputs. A 'b x m x n' tensor is interpreted as a batch of b matrices of size 'm x n'. The result is a batch of matrices computed by multiplying the corresponding pair of matrices in the two input batches. | |||
2014-11-07 | fixing small bugs in prod, cumsum, cumprod | soumith | |
2014-10-22 | Merge branch 'master' of https://github.com/torch/torch7 into clamp | Jonathan Tompson | |
2014-10-22 | added a clamp method for tensors. | Jonathan Tompson | |
2014-10-22 | Added torch.round function | Georg Ostrovski | |
2014-10-17 | Added logical any and all (similar to numpy) | Jonathan J Hunt | |
2014-06-13 | Merge pull request #42 from nicholas-leonard/master | Soumith Chintala | |
torch.renorm | |||
2014-06-11 | Merge pull request #43 from fidlej/topic_non_float_bernoulli | Soumith Chintala | |
Used double to hold the 'p' for bernoulli() and geometric() | |||
2014-06-11 | Used double to hold the 'p' for bernoulli() and geometric() instead of 'real'. | Ivo Danihelka | |
2014-06-11 | renorm only renormalizes rows with greater norms than maxnorm | nicholas-leonard | |
2014-06-11 | cwrapped torch.renorm | nicholas-leonard | |
2014-05-15 | tensor: fix squeeze(dim) behavior | Ronan Collobert | |
2014-05-09 | Merge pull request #39 from d11/fix_numel_cwrap | Ronan Collobert | |
Fix numel() cwrap return type | |||
2014-05-09 | Fix numel() return type | Dan Horgan | |
Should always be long; it was being incorrectly cast to the element type of the tensor. Also added test for nElement and numel. | |||
2014-03-20 | merged THGenerator with multinomial | Nicholas Leonard | |
2014-03-20 | Merge pull request #13 from timharley/thgenerator | Ronan Collobert | |
Make random number generator state a TH type | |||
2014-03-07 | removed debugging code. added unit tests | Nicholas Leonard | |
2014-03-06 | Reorder arguments to Tensor functions. | Tim Harley | |
2014-03-06 | multinomial with replacement | Nicholas Leonard | |
2014-03-04 | Implement method abs() for LongTensor and IntTensor | Sergio Gomez | |
2014-02-28 | Torch tests now pass! Allow all random functions to use the default ↵ | Tim Harley | |
generator if one isn't supplied when calling the function. | |||
2014-02-28 | Miraculously bulds. Very unhappy when you try to run it. | Tim Harley | |
2014-02-28 | Starting point for wrapping the mersenne state in a Lua object. | Tim Harley | |
2014-02-14 | fix dependency issues related to the new cwrap standalone package | Ronan Collobert | |
2013-08-15 | Added wrappers for Lapack functions: porti, ports, portf. | Clement Farabet | |
These are for positive symmetric matrices (inverse, solve and factorize). | |||
2013-06-14 | Added a new torch.match() function. | Clement Farabet | |
That function allows matching batches of vectors with batches of kernels. It's general enough that I thought it would be a nice addition to the base module. | |||
2013-05-10 | Added logical operators that return non-ByteTensors | Andreas Fidjeland | |
The existing equality operators always return ByteTensors. Often the return values from an equality operation needs to be used with a tensor of the original type (e.g. point-wise multiplication with a mask), and the ByteTensor therefore has to be cast back to this type. This casting is unneccessary, and for CudaTensors would be costly, as it involves a round-trip to the host. This commit adds new forms of the equality operators that directly writes the result into a tensor of the approriate type. For example x = torch.DoubleTensor(5) y = torch.rand(5) x:lt(y,0.5) The existing ByteTensor-returning forms are still present. | |||
2012-10-22 | max/min/sort: no need to increment optional IndexTensor | Ronan Collobert | |
2012-09-12 | Merge branch 'master' into noid | Ronan Collobert | |
Conflicts: extra/cuda/pkg/cutorch/TensorMath.lua | |||
2012-09-04 | Random with Cuda. | Clement Farabet | |
2012-08-13 | torch now complies with the new luaT API | Ronan Collobert | |
2012-05-06 | corrected bug in abs computation of numbers using torch.abs. | koray kavukcuoglu | |
2012-05-03 | add documentation for general eigen values and rename the functions | koray kavukcuoglu | |
eig->symeig since this is only for symmetric matrices reig->eig since this solves all cases. | |||
2012-05-03 | Add non-sym eigen value computation. | koray kavukcuoglu | |
2012-03-07 | bug correction: consider accreal in torch pkg | Ronan Collobert | |
2012-03-07 | normall/norm in TH + ported those to torch pkg | Ronan Collobert | |
2012-02-26 | add atan2 | koray kavukcuoglu | |
2012-02-21 | take beack tensor conv stuff, it is not ready yet | koray kavukcuoglu | |