Rasa NLU don't unload/delete model from project and machine

Hi !!!

Rasa NLU version: 0.14.0a1

Operating system (linux, …): Ubuntu 18.04.1 LTS

Content of model configuration file:

language: "en"

pipeline:
- name: "tokenizer_whitespace"
- name: "ner_crf"
- name: "intent_featurizer_count_vectors"
- name: "intent_classifier_tensorflow_embedding"
intent_tokenization_flag: true
intent_split_symbol: "+"

Issue:

RASA NLU server don’t delete model from my project, bellow I’m describing what I’m doing

1. run rasa server


$ python3 -m rasa_nlu.server --path /home/sal/Bot/RasaModels --max_training_processes 3  --response_log .LOGS/RASA_LOGS

2. create and train model

$ curl -XPOST  "localhost:5000/train?project=/home/sal/Bot/RasaModels&model=MODEL_EN" -d @/home/sal/Bot/trainData/train_file_MODEL_EN.json | python -mjson.tool

3. check rasa status

$ curl localhost:5000/status | python -mjson.tool

output

{
    "available_projects": {
        "/home/sal/Bot/RasaModels": {
            "available_models": [
                "fallback",
                "MODEL_EN"
            ],
            "current_training_processes": 0,
            "loaded_models": [
                "fallback"
            ],
            "status": "ready"
        },
        "default": {
            "available_models": [
                "fallback"
            ],
            "current_training_processes": 0,
            "loaded_models": [
                "fallback"
            ],
            "status": "ready"
        }
    },
    "current_training_processes": 0,
    "max_training_processes": 3
}

4. Parse me model

I get good result

$ curl -XPOST localhost:5000/parse -d '{"q":"hello there", "project": "/home/sal/Bot/RasaModels", "model": "MODEL_EN"}'

PROBLEM

I sending request to delete model


(venv) sal@name:~/Bot$ curl -XDELETE "localhost:5000/models?project=/home/sal/Bot/RasaModels&model=MODEL_EN" | python -mjson.tool

 % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100    14    0    14    0     0   2800      0 --:--:-- --:--:-- --:--:--  2800
"MODEL_EN"


RASA LOGs/output


2019-01-24 11:59:52+0100 [-] "127.0.0.1" - - [24/Jan/2019:10:59:51 +0000] "DELETE /models?project=/home/sal/Bot/RasaModels&model=MODEL_EN HTTP/1.1" 200 14 "-" "curl/7.58.0"

RASA still keep the model and don’t delete it, even when i do rasa status, still show me the model like in point 3

$ curl localhost:5000/status | python -mjson.tool

RASA delete model just from memory.


curl -X DELETE localhost:5000/models?project=my_restaurant_search_bot&model=model_XXXXXX

Hey SalAlba, im starting with version 0.14.0 of Rasa NLU (HTTP API) and im with doubts about put my pipeline in json format.

My pipeline

Can you help?

@sfurao Tell me how I can help you ?? what are your question ??

Hey SalAlba i update to 0.14.0 rasa nlu version and I’m having trouble doing the training request through htttp

i already did pip install rasa_nlu[tensorflow] and pip install tensorflow , so im not undertanding the error!

i have pyhton3.4

please print in terminal the Command bellow :

$ pip3 list

and what is your operating system on machine ?

My operating system is centOS 6

@SalAlba Did you find a solution for deleting the model from the list of available_models? I have the same issue with the latest Rasa NLU version (0.15.0a1).

1 Like