AWS Lambda: Remove unused libraries for Tensorflow 2 inference

I am trying to deploy a Lambda function on AWS, through Serverless, that requires Tensorflow 2, Pandas and Numpy;

I would like to remove from the serverless.yml file, the Tensorflow libraries not required for inference as briefly done in the following sections:

(...)
 plugins:
  - serverless-python-requirements
custom:
  pythonRequirements:
    dockerizePip: true
    zip: true
    slim: true
    useDownloadCache: true
    slimPatterns:
      - "**/tensorboard*"
      - "**/markdown*"
      - "**/werkzeug*"
      - "**/grpc*"
      - "**/tensorflow/contrib*"
      - "**/tensorflow/include*"
      - "**/external/*"

    noDeploy:
      - boto3
      - botocore
      - docutils
      - jmespath
      - pip
      - python-dateutil
      - s3transfer
      - setuptools
      - six 

Within my handler.py script, the only Tensorflow-based lines are the following:

# Import a model with Dense layers only
tf.keras.models.load_model(model_name)

# Run inference
predictions = model.predict(my_array)

However, whenever I try to deploy the service using "sls deploy", it raises an error due to file size:

Uploading service tensorflow-lambda-demo-1234.zip file to S3 (336.22 MB)...

(...)

An error occurred: HelloLambdaFunction - Unzipped size must be smaller than 262144000 bytes (Service: Lambda, Status Code: 400, Request ID: 3805f9b4-eb32-4d11-b54a-3437bf2b6d6b, Extended Request ID: null).

Which sub-libraries ("slimPatterns") and whole packages ("noDeploy") can I further exclude from the deployment package to reduce its size?