Hugging Face

Hugging Face is a platform where the machine learning community collaborates on models, datasets, and applications. Hugging Face's dedicated Inference Endpoints and Serverless Inference API allow for running any model via an HTTP REST API, which can be accessed from Pipelines using the webhook transformer.

warning

The Serverless Inference API is free but rate-limited. It should only be used for testing scenarios, and it is expected that some operations may be dropped if the specified model is not already loaded in memory. Inference Endpoints should be used in production scenarios.

The following example demonstrates using the Hugging Face Serverless Inference API to perform sentiment analysis of audio samples streamed to Golioth. The Hubert-Large for Emotion Recognition model is targeted. Sentiment analysis results are delivered to LightDB Stream as a JSON payload.

Example sentiment analysis result:

[
  {
    "score": 0.6310836672782898,
    "label": "neu"
  },
  {
    "score": 0.2573806643486023,
    "label": "sad"
  },
  {
    "score": 0.09393830597400665,
    "label": "hap"
  },
  {
    "score": 0.017597444355487823,
    "label": "ang"
  }
]

Reminder

Make sure to create a secret named HUGGING_FACE_TOKEN in the format Bearer <token>.

Pipeline: Hugging Face Emotion Recognition

filter:
  path: "/audio"
steps:
  - name: emotion-recognition
    transformer:
      type: webhook
      version: v1
      parameters:
        url: https://api-inference.huggingface.co/models/superb/hubert-large-superb-er
        headers:
          Authorization: $HUGGING_FACE_TOKEN
  - name: embed
    transformer:
      type: embed-in-json
      version: v1
      parameters:
        key: text
  - name: send-lightdb-stream
    destination:
      type: lightdb-stream
      version: v1