Watson NLP custom model

This project contains assets and information related to create a custom Watson NLP for Embed models.

Related blog post Create a custom dictionary model for Watson NLP and Watson NLP for Embed customize a classification model.

• Customize a classification model for Watson NLP for Embed
• Run Watson NLP for Embed with a customized model on the local computer

Visit the list of the supported models for Watson NLP for Embed

Customize a classification model for Watson NLP for Embed

• Step 1: Familiarize yourself with Watson Studio
• Step 2: Clone the project to your local computer
• Step 3: Open your Watson Studio project on IBM Cloud
• Step 4: Select New asset
• Step 5: Press Jupyther notebook editor
• Step 6: Select runtime Jupyther notebook editor
• Step 7: Upload the Jupyther notebook file watson-nlp-custom-model-for-text-classification.ipynb from your local machine and press Create
• Step 8: Execute the steps in the Jupyter notebook

Step 1: Familiarize yourself with Watson Studio

Follow the steps of the blog post Customize a model for Watson NLP to get familiar with the Watson Studio.

Step 2: Clone the project to your local computer

git clone https://https://github.com/thomassuedbroecker/watson-nlp-custom-model.git
cd watson-nlp-custom-model

Step 3: Open your Watson Studio project on IBM Cloud

Step 4: Select New asset

Step 5: Press Jupyther notebook editor

Step 6: Select runtime Jupyther notebook editor

• Runtime: DO + NLP Runtime 22.1 on Python 3.9 (2 CPU 8 GB RAM)

Step 7: Upload the Jupyther notebook file watson-nlp-custom-model-for-text-classification.ipynb from your local machine and press Create

Location of the file YOUR_PATH/watson-nlp-custom-model/juypter-notebook/watson-nlp-custom-model-for-text-classification.ipynb.

Step 8: Execute the steps in the Jupyter notebook

Link to the notebook: Watson NLP custom model for text classification.

Run Watson NLP for Embed with a customized model on the local computer

Note: Your need to create a customized model and save it in the docker/app/models folder. Please read the readme.

• Step 1: Clone the repository
• Step 2: Navigate to the docker folder
• Step 3: Create a .env file
• Step 4: Add an IBM Cloud access key to your local .env file
• Step 5: Execute the bash automation
• Step 6: Open a new terminal and insert following curl command

Step 1: Clone the repository

git clone https://https://github.com/thomassuedbroecker/watson-nlp-custom-model.git
cd watson-nlp-custom-model

Step 2: Navigate to the docker folder

cd docker

Step 3: Create a .env file

cat .env_template > .env

Step 4: Add an IBM Cloud access key to your local .env file

nano .env

Content of the file:

export IBM_ENTITLEMENT_KEY="YOUR_KEY"
export IMAGE_REGISTRY="cp.icr.io/cp/ai"
export RUNTIME_IMAGE="watson-nlp-runtime:1.0.20"
export WATSON_RUNTIME_BASE="$IMAGE_REGISTRY/$RUNTIME_IMAGE"

Step 5: Ensure you have copied your model to the folder docker/app/models.

Please read the readme.

Step 5: Execute the bash automation

The creation can take up to 10 minutes for the download of the runtime image.

sh run-watson-nlp-with-docker.sh

• Example output:

# ******
# Connect to IBM Cloud Container Image Registry: cp.icr.io/cp/ai
# ******

IBM_ENTITLEMENT_KEY: 
...

Login Succeeded

# ******
# Build container image locally
# Runtime container image: cp.icr.io/cp/ai/watson-nlp-runtime:1.0.20
# ******
[+] Building 0.2s (7/7) FINISHED                                  
 => [internal] load build definition from 
 ...
 
# ******
# Run NLP
# ******

# Run the runtime with the models mounted

[STARTING RUNTIME]
...
"2023-01-25T20:25:07.040114"}
[STARTING GATEWAY]
2023/01/25 20:25:07 Running with INSECURE credentials
2023/01/25 20:25:07 Serving proxy calls INSECURE

Step 6: Open a new terminal and insert following curl command

curl -s -X POST "http://localhost:8080/v1/watson.runtime.nlp.v1/NlpService/ClassificationPredict" \
  -H "accept: application/json" \
  -H "grpc-metadata-mm-model-id: ensemble_model" \
  -H "content-type: application/json" \
  -d "{ \"rawDocument\": \
    { \"text\": \"The credit card doesn t work, and I look at the savings, but I need more money to spend.\" }}" | jq

• Example output:

{
  "classes": [
    {
      "className": "Credit card or prepaid card",
      "confidence": 0.5404788
    },
    {
      "className": "Debt collection",
      "confidence": 0.22745371
    },
    {
      "className": "Credit reporting, credit repair services, or other personal consumer reports",
      "confidence": 0.22019744
    },
    {
      "className": "Checking or savings account",
      "confidence": 0.100063786
    },
    {
      "className": "Mortgage",
      "confidence": 0.043992974
    }
  ],
  "producerId": {
    "name": "Voting based Ensemble",
    "version": "0.0.1"
  }
}