search

zhongkaifu / RNNSharp

RNNSharp is a toolkit of deep recurrent neural network which is widely used for many different kinds of tasks, such as sequence labeling, sequence-to-sequence and so on. It's written by C# language and based on .NET framework 4.6 or above versions. RNNSharp supports many different types of networks, such as forward and bi-directional network, sequence-to-sequence network, and different types of layers, such as LSTM, Softmax, sampled Softmax and others.
BSD 3-Clause "New" or "Revised" License
285 stars 92 forks source link

maybe it's time to support CUDA ! #36

Closed AvenSun closed 7 years ago

AvenSun commented 7 years ago

Hi, RNNSharp is great for .net developers. but sometimes training time is kinda long. I found one interesting project managedCuda which is complete .net wrapper for the CUDA Driver API. so maybe it's a good idea to merge it into rnnsharp. best regards! Adesun

zhongkaifu commented 7 years ago

Thanks @Adesun

I once slightly modified RNNSharp to apply managedCuda, however, its performance is worse than CPU+SIMD version. :(

To leverage GPU performance completely, RNNSharp has to have a large design and code modification. It's timing cost and currently, I don't have bandwidth for it.

AvenSun commented 7 years ago

I see. thx.