Starter Pack Problems

juste

(Ted Cohn) #1

I’m a noob! Trying to get the starter pack working on a Mac. Preface: I’m a Python noob too. I’ve gotten pretty far with the stock repo, but noted some problems and can’t get past the server and cmdline tasks:

  1. I installed everything suggested that I’m aware of according to the online doc but saw many red lines saying items weren’t installed and should be, but the installation process succeeded, so should I be concerned?

  2. The quickstart doc seems out of date and doesn’t match the starter pack README.md file which has make commands that differ from the doc.

  3. I was forced to run pip install dask --upgrade in order to train the core. Wasn’t sure why this wasn’t already installed per rasa installation scripts.

  4. When I run make cmdline. How can I get rid of the warning?

/Users/ted/anaconda3/lib/python3.6/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
  from ._conv import register_converters as _register_converters
2019-03-13 09:39:03 INFO     root  - Rasa process starting
  1. When I run make cmdline, it takes over a minute to load the bot and start accepting input. Is this normal? What can be done to accelerate this for test/debug experience?

  2. Is it possible to test the interaction of the agent in a single Python instance from the command line instead of starting a server?

Thanks!


(Juste) #2

Hey @ted. Let’s see if I can help with any of your questions:

  1. Can you share the log of the installation (the part where you are told that some dependencies are not installed) so that we could look into it?
  2. Quckstart wasn’t intended to match the starterpack (it’s just because starterpack came much later than the quickstart), but I see your point of having a unified example and we will definitely take this into consideration.
  3. Might be related to 1 so it would be super helpful to see which dependencies failed to install.
  4. One way to get rid of the warning is to shut them down using warnings lib. I’ll look into it as well.
  5. make cmdline shouldn’t take this long. You could add --debug flag to the training command when your train nlu and core model to see if there is anything odd going on.
  6. You can test the NLU model from python, but to test the core model you do have to start the server. Any particular reason why you would prefer not using the server?