Hi @TheCedarPrince. I ran the script on freshly installed Raspian Buster but wasn’t able to run Rasa because it couldn’t find TensorFlow models. Do you know what would be causing this?
Hey @zshoaib - not entirely sure what the issue is but it appears you have an error here:
Should the name be “MemorizationPolicy” and not “MemoriationPolicy”? Seems like a typo.
Hi @TheCedarPrince. No, Memoization policy is a part of RASA policies: Policies
On second look, this looks a bit more involved than I initially thought. It would appear there is an issue with tensor2tensor - I will have to take a look at this more later. Right now, I am working with @stephens on developing a Dockerfile for easy installation - please check back a bit later. Will let you know what happens.
I’ve created an install script that you can see-get here. It does a clean install of the latest Rasa v1.60 on the current raspbian-buster(2019-09-26).
Mine is different from what others have posted earlier in the thread. It works fairly quickly and creates a clean install and allows Rasa models to run well from my testing on a Raspberry Pi4 or 3B+. But it fails in training mode! For my use-case it works…I train the model on a different computer and move it to the Pi where it does just fine.
I’d love to hear from others who might have a better handle as to trainning issues.
1 Like
Hi, it doesn’t work on my Pi 4 4Gb with Ubuntu server 19.10 64 bits
@TheCedarPrince @stephens @ziligy Thank you for all your work, I have trained a Rasa model on ubuntu PC using Rasa 1.7.0 and its working great,
But I wish to transfer the trained model from my Ubuntu to Raspberry Pi and then run on it,
- What are the steps I should follow to achieve above?
- Is the above scripts compatible with models trained using Rasa 1.7.0?
- I do not wish to train the model on Raspberry Pi but only run the model so which Raspberry Pi model I better invest on?
I am eagerly waiting for your answer, Thank you
I don’t think we can install Rasa on Raspberry 
@groot I was able to do an install of Rasa on Raspbian based on the script from TheCedarPrince. I think you can see that above. I also have a repo at [GitHub - rgstephens/rasaPi: Rasa with Docker on a Raspberry Pi] which includes this script.
I have made some attempts to create a docker ARM image without any success (you can also see this in the repo above). @JulianGerhard21 has also put some time into making a docker image.
Some of us are thinking that it’s better to wait and try again for a docker image when a TensorFlow 2 based Rasa is released.
1 Like
I’m using 64-bit Ubuntu on my raspberry 4 4gb. I think that’s why I can’t install Rasa.
Thank you for your reply! Any timeline when TensorFlow 2 based on Rasa will be released?
Did you try @ziligy method mentioned here Running RASA on the RPi 4 with Raspbian Buster! he says he was successful on running a Rasa model on raspberry pi using his script here shell script to install rasa v1.6.0 on raspberry pi raspbian/buster · GitHub, he used raspbian-buster instead of ubuntu Let me know how it goes if you attempt it
I’ve already tried this script along with a script named install.sh found in another ticket on github.
Both scripts lead to a tensorflow 
installation error
1 Like
I trying to build tensor2tensor, spacy from source but n ever succeeded, Please if someone can provide whl of tensor2tensor=1.14 for python3.6.8
Hi guys,
just to give a small update - I finally managed to install Rasa 1.8 with all its dependencies on Debian Buster on a Raspberry Pi 4.
I will soon update the script provided.
Kind regards
Julian
6 Likes
Hi, I am trying to do this from a long time, can you share the method? Thanks
Hi @divyanshu,
can you please try it with this tutorial and provide me with some feedback?
Regards
Julian
README_18.md (5.5 KB)
Hi @JulianGerhard, thanks for this walkthrough, I got rasa installed successfully, but while installing spacy I got these errors. errorspacy (19.3 KB)
Hi @divyanshu,
ok - I will have a look into it.
Did you respect export BLIS_ARCH=generic before installing?
Kind regards Julian
Hi, @JulianGerhard yep, I am sure of it. Also i got error in rasa while selecting models. In “rasa shell nlu” it was working well but in program on line nlu_model = Interpreter.load(’./models/nlu’) it throws this error.
(rasa_env) pi@raspberrypi:~/voicebot $ LD_PRELOAD=/usr/lib/arm-linux-gnueabihf/libatomic.so.1 python main.py /home/pi/rasa/rasa_nlu/init.py:12: UserWarning: The ‘rasa_nlu’ package has been renamed. You should change your imports to use ‘rasa.nlu’ instead. UserWarning, WARNING:tensorflow:From /home/pi/rasa_env/lib/python3.7/site-packages/tensorflow_core/python/ops/array_grad.py:563: _EagerTensorBase.cpu (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version. Instructions for updating: Use tf.identity instead. 2020-03-17 14:44:14.776869: W tensorflow/core/kernels/data/generator_dataset_op.cc:103] Error occurred when finalizing GeneratorDataset iterator: Cancelled: Operation was cancelled
1 Like