can I use multi-gpu for training ?

meng-tang / rloss

Regularized Losses (rloss) for Weakly-supervised CNN Segmentation

MIT License

208 stars 46 forks source link

Open yaoqi-zd opened 5 years ago

yaoqi-zd commented 5 years ago

Hi, can I use multi-gpu for training ? since the densecrf loss layer is implemented in C++, I'm afraid that only single-gpu training is support ?