Rasa without a GPU

post by Val on Nov 28, 2019

Hello,

I am trying to make Rasa work with a 2.1 capacity GPU (outside of the tensorflow_gpu minimum requirement). Is it possible to use Rasa only with tensorflow CPU?

The error log that blocked me all day is the following :

$ rasa init
Welcome to Rasa! 🤖

To get started quickly, an initial project will be created.
If you need some help, check out the documentation at https://rasa.com/docs/rasa.
Now let's start! 👇🏽

? Please enter a path where the project will be created [default: current directory] ./rasa_projects
? Directory '/home/XXX/rasa_projects' is not empty. Continue?  Yes
Created project directory at '/home/XXX/rasa_projects'.
Finished creating project structure.
Training an initial model...
Training Core model...
2019-11-28 16:47:12 INFO     numexpr.utils  - NumExpr defaulting to 4 threads.
2019-11-28 16:47:21 INFO     absl  - Entry Point [tensor2tensor.envs.tic_tac_toe_env:TicTacToeEnv] registered with id [T2TEnv-TicTacToeEnv-v0]
Processed Story Blocks: 100%|███████| 5/5 [00:00<00:00, 2925.71it/s, # trackers=1]
Processed Story Blocks: 100%|████████| 5/5 [00:00<00:00, 753.88it/s, # trackers=5]
Processed Story Blocks: 100%|███████| 5/5 [00:00<00:00, 251.17it/s, # trackers=20]
Processed Story Blocks: 100%|███████| 5/5 [00:00<00:00, 168.49it/s, # trackers=24]
Processed trackers: 100%|███████████| 5/5 [00:00<00:00, 2252.10it/s, # actions=16]
Processed actions: 16it [00:00, 7839.82it/s, # examples=16]
Processed trackers: 100%|███████| 231/231 [00:00<00:00, 869.31it/s, # actions=126]
2019-11-28 16:47:23.292187: F tensorflow/stream_executor/lib/statusor.cc:34] Attempting to fetch value instead of handling error Internal: no supported devices found for platform CUDA
Abandon (core dumped)

post by Tanja on Dec 2, 2019

We are training our models on GPUs from time to time without any problems. What Rasa version are you using?

post by Val on Dec 3, 2019

I’m using the most recent version, the error occurred just after my rasa download with the $rasa init. My question is: is it possible to use Rasa without a GPU in your machine? Or in this case a GPU that doesn’t support Tensorflow?

The error message I got :

“no supported devices found for platform CUDA Abandon (core dumped)”

Tells me it can’t and I tried using manual installation and other guides accessible on your documentation. Is there a way to use RASA using only tensorflow-CPU?

post by dakshvar22 on Dec 4, 2019

@Val Have you installed tensorflow-gpu on your machine? If you wish to use CPU only, just install tensorflow==1.15.0, that should work. In case you have both tensorflow and tensorflow-gpu installed and you know that your GPU is incompatible with tensorflow-gpu please uninstall tensorflow-gpu. Let us know if you face any further problem.

post by Val on Dec 4, 2019

Problem solved! thanks a lot!

post by olegtropinin on Jan 26, 2020

It’s another way to solve it. Need to remove tensorflow and install tensorflow-cpu.

pip uninstall tensorflow
pip install tensorflow-cpu