OpenCL backend for Torch nn neural networks library.
Please see distro-cl for installation instructions.
- nn.Linear
These mostly 'just work', since based on underlying tensor methods, already implemented in cltorch. Tested with:
- nn.Narrow
- nn.Identity
- nn.Dropout
- nn.SpatialConvolutionMM
- nn.SpatialMaxPooling (including
ceil
mode) - nn.SpatialAveragePooling
- nn.TemporalConvolution2 This is specific to clnn. It works on cpu and cuda too, not just on OpenCL. It is API-compatible with TemporalConvolution, and faster than TemporalConvolution, on both CUDA and OpenCL.
- nn.Tanh
- nn.Sigmoid
- nn.ReLU
- nn.ELU
- nn.Exp
- nn.Sqrt
- nn.Square
- nn.Abs
- nn.LogSigmoid
- nn.HardTanh
- nn.LogSoftMax
- nn.SoftMax (including spatial mode)
These 'just work', since they are based on underlying torch operations, which are already implemented in cltorch. Tested with:
- nn.CMulTable
- nn.CAddTable
- nn.MSECriterion
- nn.ClassNLLCriterion
Containers 'just work', since they just call standard operations on the contained modules. Tested with:
- nn.Sequential
- nngraph
In theory, trainers 'just work', since they just call standard torch methods on the network. The following are good first choices:
- nn.StochasticGradient
- optim.lbfgs
- optim.adam
Please see https://github.com/soumith/convnet-benchmarks#imagenet-winners-benchmarking
- On a Titan X, OpenCL torch is about 3 times slower than CUDA torch
- eg for VGG, cutorch takes 1100ms, and cltorch takes 3400ms
- Andrej's char-rnn is OpenCL-enabled, simple add option
-opencl 1
- Justin's neural-style has an OpenCL port in progress by Shubhanshu napsternxg/neural-style
Porting guidelines, for project maintainers, available here: porting-guidelines.md.
- 2nd May:
- Re-applied:
- 26th March:
- add TemporalConvolution2: same API and usage as TemporalConvolution, but faster on GPUs
- 26th March:
- Re-applied:
- 31st April:
- 30th April:
- rolled back to as-of 21st February, prior to lots of THNN changes in upstream Torch
- additionally, installation procedure is now to use a specific torch distro, for stability
- 1st Feb:
- merged/ported THNN phase 3. Any weird build issues, please update both
nn
andclnn
.
- merged/ported THNN phase 3. Any weird build issues, please update both
- 2nd January, 2016:
- merged/ported THNN architecture across, and the implementation of Abs, so the unit-tests pass again now
- 15th December:
- merged Sergey's SpatialAveragePadding and ceil kernels into
master
branch
- merged Sergey's SpatialAveragePadding and ceil kernels into
- 29th November:
- added ELU
- 25th September:
- ported Sergey's not-yet-merged SpatialAveragePadding and ceil kernels, into
clnn-avgpool
branch - ported latest version of SoftMax, ie essentially Jonghoon's Update SoftMax to work in spatial mode
- ported Sergey's not-yet-merged SpatialAveragePadding and ceil kernels, into
- 23rd September:
- ported latest cunn implementation of
SpatialMaxPooling
across, ie approximately Sergey's Deterministic max-pooling PR- this includes
:ceil()
implementation
- this includes
- ported latest cunn implementation of
- 22nd September:
- added non-batch implementation of LogSoftMax (previously only handled batched input)
- added SoftMax, for both batched and non-batched
- 20th September:
- added non-batch implementation for SpatialMaxPooling (previously only handled batched input), for contiguous pools