Will the CONV operation be released in future?

[iamweiweishi]

Thank you for your great contribution. I noticed that the dense layer could be incorporated in VGG successfully, but I did not found the implementation of CONV layer, I would like to know if the CONV layer will be implemented in the near future?

[JunyaoPu]

Hello! I am new to this library. I am also wondering that does the T3F library has Conv layer support?

[Bihaqo]

Sorry for an embarrassingly long reply. As this library focuses on a particular tensor decomposition (Tensor Train decomposition) and as I don't know a good way of using TT-decomposition for CONV layers*, I deprioritised this and probably won't have time to implement it anytime soon.

If you don't mind using pytorch, check out TensorLy-torch though, it has a lot of layers including TT-linear layer and different factorized CONV layers (CP, Tucker, and TT, so you can try different ones for yourself and choose the best one).

* I contributed to this paper about a TT-CONV layer, but it's not actually very good (hinted by the fact that it's not published anywhere but arxiv :) )

[JunyaoPu]

Sorry for an embarrassingly long reply. As this library focuses on a particular tensor decomposition (Tensor Train decomposition) and as I don't know a good way of using TT-decomposition for CONV layers*, I deprioritised this and probably won't have time to implement it anytime soon.

If you don't mind using pytorch, check out TensorLy-torch though, it has a lot of layers including TT-linear layer and different factorized CONV layers (CP, Tucker, and TT, so you can try different ones for yourself and choose the best one).

• I contributed to this paper about a TT-CONV layer, but it's not actually very good (hinted by the fact that it's not published anywhere but arxiv :) )

Thank you for your reply! I appreciate your T3F library, It really helps me to implement my work with TT format on Tensorflow, and thanks for sharing the TensorLy-torch I will talk a look at it.