Step 1: Import and Instantiate a Workflow

Import the HuggingfaceInferenceClientWorkflow class from infernet_ml and create an instance of it.

from infernet_ml.workflows.inference.hf_inference_client_workflow import HuggingfaceInferenceClientWorkflow

In this instance, we're going to use the text_classification task type, and use the Kaludi/Reviews-Sentiment-Analysis (opens in a new tab) model. You can use any other model tagged as Text Classification (opens in a new tab) from the Huggingface model hub.

workflow = HFInferenceClientWorkflow(task="text_classification", model="Kaludi/Reviews-Sentiment-Analysis")

Step 2: Setup the Workflow

We're going to setup our model. Depending on the workflow, this does various tasks to make the model ready for inference:

• For workflows that execute the model themselves, this might do something like downloading the model weights.
• For workflows that use a remote inference service, this might setup the connection to the service, and ensure the model is available on the service.

workflow.setup()

Step 3: Perform Inference

Now we can perform inference on our model. All of the workflows in infernet-ml have a inference() method that takes in the input data and returns the output.

results = workflow.inference({
    "text": "I was extremely disappointed with this product."
})

Step 4: Putting it All Together

Finally, we can display the results of our inference. In the case of Kaludi/Reviews-Sentiment-Analysis (opens in a new tab) we expect the output to have different classes and their probabilities.

from infernet_ml.workflows.inference.hf_inference_client_workflow import HuggingfaceInferenceClientWorkflow
 
workflow = HFInferenceClientWorkflow(task="text_classification", model="Kaludi/Reviews-Sentiment-Analysis")
workflow.setup()
results = workflow.inference({
    "text": "I was extremely disappointed with this product."
})
print(f"results: {results}")

Running this code, we'll get an output similar to the following:

results: {'output': [{'label': 'Negative', 'score': 0.9863545298576355}, {'label': 'Positive', 'score': 0.013645444996654987}]}

And just like that, we've performed inference on a Huggingface model using infernet-ml!