I am currently working with Rasa 2.8.1 and i am training my system on CPU, I have tried to train it with GPU (tensorflow-gpu) but it always end up with an error. I have tried every version of tensorflow gpu. So my question is can we use rasa 2’s any version with tensorflow-gpu ? kindly help me out here.
It should be possible to train on GPU but would require a few manual steps. Someone from our community wrote a guide to help with that. We also have an open ticket to simplify this, but it’s unclear when it’s going to be done at this point. I hope this helps!
What kind of an error are you getting?
If you’re running it under Docker and it’s something like E tensorflow/stream_executor/cuda/cuda_driver.cc:314] failed call to cuInit: UNKNOWN ERROR (-1) , you need to make sure that you’re passing in
--gpus all in the Docker command, otherwise Docker won’t use the GPUs.
I am not running under docker, i just want to run my project normally but with the tensorflow-gpu support.