Every time i train the bot then train model are store in the ‘/tmp/’ folder and i have to go to the tmp folder and delete all the file . There is any way in which train model not store in tmp folder or way to delete all the trained model
Could you please tell more details? What’s your deployment? Is it
Rasa Open Source, or
Also, if that’s something you’d suggest to improve, feel free to open an feature request here
@degiz when i train my bot its model also store in my tmp folder of the system(by default settings )
I’m not sure I understand what the issue is, but for
rasa train command there’s an option:
–out OUT | Directory where your models should be stored. (default: models)
you can use it to control where the trained model will be saved.
@degiz i know rasa train command. My question is when i am run rasa train then one model is store in my default path i.e. inside the models folder and another train folder is store in my laptop temporary memory i.e tmp folder
I see. From the implementation perspective I know that we should remove anything created while training in the
/tmp folder. If that doesn’t happen on your deployment then it’s a bug. Could you please explain more about your deployment? Operating system? Is it just
rasa pypi package?
I’m facing the same issue:
After a training form within Rasa X, Rasa seems not to clean up temporary files.
amongst them, the model created by this training. See f.e. attached listing
Notice on the other hand, that training from the command line, “rasa train” does clean its temperary files.
Rasa X started from bash command line, i.e. “rasa x”.
Remark: I’ve just noticed that not only the training from within rasa x doesn’t clean the files.
Also starting rasa x from the command line creates temporary files but won’t delete them .
And there are also some temporary folders not deleted after the training
rasa_x_tmp_files.txt (414 Bytes)
same issue, after a model is deteled in /models folder to train a new one, a file in /tmp persists:
rasa.core.agent - Persisted model to ‘/tmp/tmpqakto7_q/core’