Rasa Train - Nothing changed. You can use the old model

Getting Started with Rasa
1

I installed Rasa on windows on conda env. Path of my nlu.md on my pc is C:\Users\pooja\Anaconda3\envs\rasa4\lib\site-packages\rasa\cli\initial_project\data

i edited nlu, domain and stories files and saved. Yes these are saved. and now i train it on cmd line (rasa4) C:\Users\pooja>rasa train Nothing changed. You can use the old model stored at ‘C:\Users\pooja\models\20201002-135030.tar.gz’.

running rasa shell also indicate that it did not pick changes i made. I again checked back and found files are saved.

Am i running it incorrectly? or path i am updating files are incorrect? i am sure i might be doing some silly mistake. Please suggest.

Thanks

2

This happened with me once or twice. I simply deleted all existing models from the “models” directory and then retrained the model.

Catch : It takes a longer training time since the model is trained from the scratch.

1 Like
3

Hi Gaurav, Thanks for your input. As you suggested, i deleted model zip and its trained again without any error. Now while i am doing “rasa shell” its not responding as per my inputs and behave like its reading old file.

Your input -> hi Hey! How are you? Your input -> how can i lower my stress level? I am a bot, powered by Rasa. Your input ->

Its seems its picking file from some other location and i am editing at some other. But there is only that nlu.md file in my system ( i did system search)

Any idea.

Thanks

4

can you please share relevant files? Having a look at the code will help me get a better understanding of your problem.

1 Like
5

nlu.md (1.5 KB) stories.md (590 Bytes) domain.yml (1.1 KB)

attaching 3 files where i added intent stories and responces.

Is there a command to check path of nlu file its is picking up to train during rasa train? just to confirm that it is picking the file i am updating… even though there is only 1 nlu as its my first project after installation.

6

that happend to me early but I fix it just deleting all my models. in windows I recommed that you use visual studio code to update your files and use the visual studio terminal

7

yes i am using visual studio to edit code and deleted model but again it created the same model without adding new intents neither it gave any error about intents-

How to check that the file it picked from train is same as the file i am editing? is there any command to check path of file its picking?

8

I don’t think so, but only read and detect the files inside the proyect

9

I downloaded your files and trained a model on my system.

It seems to be working as expected.

del_asap
del_asap1358×144 8.09 KB

1 Like
10

i feel i am updating files at wrong path.

i installed rasa here at conda virtual env and edited files here only - C:\Users\pooja\Anaconda3\envs\rasa4\lib\site-packages\rasa\cli\initial_project\data

But looks like rasa train is picking from somewhere else. I search system with nlu.md and only this location is coming.

or may be in cmd i am running at wrong path?

Not sure… but looks lik eits not picking my file… how to check?

11

I’ll write down steps, you can see where you’re doing it differently

  1. Open Conda command prompt
  2. Make a directory (where you want to store your bot) using mkdir command
  3. Get Inside that directory using cd command
  4. Activate your virtual environment using conda activate env_name
  5. Run the command rasa init. (It will create all the files needed).
  6. Open the directory and replace existing domain.yml, stories.yml and nlu.md with your files.
  7. Run rasa train in conda command prompt
  8. Run rasa shell to chat with your bot.

Brownie point: In your existing model, run rasa shell --debug to analyze everything

1 Like