HI, in my Tensorflow 2.6 environment all works great; all my models run on GPU without problems; when I install Rasa, it downgrade TF to 2.3 version;:(( but now, when I test for gpu available – tf.config.list_physical_devices(‘GPU’)-- TF detect my GPU anymore; then nlu, core training works only in CPU mode…not on GPU…this is very irksome… probably cuda and cudnn toolkit incompatibility ?(i think so…)…ther’s another way to turn around this problem ? because isn’t very immediate reset all compatibilities … thanks in adv
HI!
yes , this is an virtual env…the problem is that for a older TF versions, we need a specific cuda, specific cudnn e visual studio…you can’t install cuda on virtual env, then You need to handle different versions on your system…this isn’t very efficient way to go…strange that Rasa works with this version…also because, probably, when for ex, You fine tuning an external transformer model, or any external models, will be compatibily issue when it will be used in Rasa…(probably…I don’t know still Rasa framework, but I work daily with trasnformers, and RNNs…)
however if this is the Rasa way…I’ll try to adapt 
1 Like
Ah I see what you mean
Yes you have to work with what Rasa gives you, but I don’t know if there are workarounds. The time I tried to use CUDA I had errors and never looked back.