Hello, We would like to take advantage of using the two available GPUs using theano I would like to train a model in one GPU and other model in the other GPU using different scripts Until now we only use one GPU so we are wasting half of the computational capacity! I see that the GPUS are GPUS="/dev/nvidia2;/dev/nvidia3" But I would like to know what are the theano flags to select the gpu ( I have never had a multiple gpu computer) I have tried something like: THEANO_FLAGS='device=gpu0' python -c "import theano" 2> /modelState/gpu0.txt & THEANO_FLAGS='device=gpu1' python -c "import theano" 2> /modelState/gpu1.txt & THEANO_FLAGS='device=gpu2' python -c "import theano" 2> /modelState/gpu2.txt & THEANO_FLAGS='device=cuda0' python -c "import theano" 2> /modelState/cuda0.txt & THEANO_FLAGS='device=cuda1' python -c "import theano" 2> /modelState/cuda1.txt & THEANO_FLAGS='device=cuda2' python -c "import theano" 2> /modelState/cuda2.txt & But I can only get one gpu... what is the correct device to select the required gnu? Thanks!!