TensorFuse – Common interface for Theano, CGT, and TensorFlow
1–8 of 8 posts
Re: TensorFuse – Common interface for Theano, CGT, and TensorFlow
#2It would have been cool if they posted benchmarks to the different interfaces.
But for people wondering, it appears to be a Theano API for interfacing with Tensorflow, and a few other tensor/deep learning libraries.
Pretty cool idea.
Re: TensorFuse – Common interface for Theano, CGT, and TensorFlow
#3Keras (http://keras.io) is a similar library, although only for TensorFlow and Theano...
Re: TensorFuse – Common interface for Theano, CGT, and TensorFlow
#4Keras ( http://keras.io ) is a similar library, although only for TensorFlow and Theano...
Keras is a very high-level library compare to tensorfuse. Tensorfuse is primarily built to easy migration from theano to something else.
Re: TensorFuse – Common interface for Theano, CGT, and TensorFlow
#5JQuery for scientific frameworks?
Re: TensorFuse – Common interface for Theano, CGT, and TensorFlow
#6Keras ( http://keras.io ) is a similar library, although only for TensorFlow and Theano...
Keras is a very high-level library compare to tensorfuse. Tensorfuse is primarily built to easy migration from theano to something else.
Yeah, but Keras has a lower level abstraction layer to switch between Theano and Tensorflow back ends.
Re: TensorFuse – Common interface for Theano, CGT, and TensorFlow
#7Keras ( http://keras.io ) is a similar library, although only for TensorFlow and Theano...
Keras is a very high-level library compare to tensorfuse. Tensorfuse is primarily built to easy migration from theano to something else.
Keras has a low-level compatibility library[1] (`from keras import backend as K`) that people have reported as useful independently.
The interface offered by K[2] seems quite similar to that of TensorFuse[3]
As the K backend is used for Keras, it also offers proof that K can be used for sizable and complex projects, plus can take advantage of pre-existing testing for Keras.
Whilst I love the idea of CGT, it has not yet taken off. I'd be far more interested in seeing a Neon[4] backend considering it has the fastest performance across the board on existing hardware and they're planning to release their own hardware soon.
[2]: https://github.com/fchollet/keras/blob/master/keras/backend/...
[3]: https://github.com/dementrock/tensorfuse/tree/master/tensorf...
[4]: https://github.com/NervanaSystems/neon
Edit: Incorrectly thought TensorFuse didn't support RNNs, thanks dementrock! Also excited that Lasagne has someone working on being backend independent!
Re: TensorFuse – Common interface for Theano, CGT, and TensorFlow
#8Earlier quoted context omitted.
Keras is a very high-level library compare to tensorfuse. Tensorfuse is primarily built to easy migration from theano to something else.
Keras has a low-level compatibility library[1] (`from keras import backend as K`) that people have reported as useful independently. The interface offered by K[2] seems quite similar to that of TensorFuse[3] As the K backend is used for Keras, it also offers proof that K can be used for sizable and complex projects, plus can take advantage of pre-existing testing for Keras. Whilst I love the idea of CGT, it has not y…
Actually TensorFuse does support RNN by porting a subset of scan to TensorFlow. I've been using it to port Lasagne to support TensorFlow: https://github.com/dementrock/Lasagne-tf. The examples/recurrent.py there actually works.
Good suggestion on Neon!