app.py

from deepsparse import Pipeline
import time
import gradio as gr

markdownn = '''
# Named Entity Recognition Pipeline with DeepSparse
Named Entity Recognition is the task of extracting and locating named entities in a sentence. The entities include, people's names, location, organizations, etc. 
![Named Entity Recognition Pipeline with DeepSparse](https://huggingface.co/spaces/neuralmagic/nlp-ner/resolve/main/named.png)

## What is DeepSparse?
DeepSparse is an inference runtime offering GPU-class performance on CPUs and APIs to integrate ML into your application. Sparsification is a powerful technique for optimizing models for inference, reducing the compute needed with a limited accuracy tradeoff. DeepSparse is designed to take advantage of model sparsity, enabling you to deploy models with the flexibility and scalability of software on commodity CPUs with the best-in-class performance of hardware accelerators, enabling you to standardize operations and reduce infrastructure costs.
Similar to Hugging Face, DeepSparse provides off-the-shelf pipelines for computer vision and NLP that wrap the model with proper pre- and post-processing to run performantly on CPUs by using sparse models.

SparseML Named Entity Recognition Pipelines integrate with Hugging Face’s Transformers library to enable the sparsification of a large set of transformers models. 
### Inference API Example
Here is sample code for a token classification pipeline: 
```python
from deepsparse import Pipeline
pipeline = Pipeline.create(task="ner", model_path="zoo:nlp/token_classification/distilbert-none/pytorch/huggingface/conll2003/pruned80_quant-none-vnni")
text = "Mary is flying from Nairobi to New York"
inference = pipeline(text)
print(inference)
```
## Use Case Description
The Named Entity Recognition Pipeline can process text before storing the information in a database. 
For example, you may want to process text and store the entities in different columns depending on the entity type. 

[Want to train a sparse model on your data? Checkout the documentation on sparse transfer learning](https://docs.neuralmagic.com/use-cases/natural-language-processing/question-answering)
'''
task = "ner"
sparse_qa_pipeline = Pipeline.create(
        task=task,
        model_path="zoo:distilbert-conll2003_wikipedia_bookcorpus-pruned90",
    )

def map_ner(inference):
  entities = []
  for item in dict(inference)['predictions'][0]:
    dictionary = dict(item)
    entity = dictionary['entity']
    if entity == "LABEL_0":
        value = "O"
    elif entity == "LABEL_1":
        value = "B-PER"
    elif entity == "LABEL_2":
        value = "I-PER"
    elif entity == "LABEL_3":
        value = "-ORG"
    elif entity == "LABEL_4":
        value = "I-ORG"
    elif entity == "LABEL_5":
        value = "B-LOC"
    elif entity == "LABEL_6":
        value = "I-LOC"
    elif entity == "LABEL_7":
        value = "B-MISC"
    else:
        value = "I-MISC"
    dictionary['entity'] = value
    entities.append(dictionary)
  return entities
      
def run_pipeline(text):
    sparse_start = time.perf_counter()
    sparse_output = sparse_qa_pipeline(text)
    sparse_entities = map_ner(sparse_output)
    sparse_output = {"text": text, "entities": sparse_entities}
    sparse_result = dict(sparse_output)
    sparse_end = time.perf_counter()
    sparse_duration = (sparse_end - sparse_start) * 1000.0
    
    return sparse_output, sparse_duration


with gr.Blocks() as demo:       
    with gr.Row():
        with gr.Column():
            gr.Markdown(markdownn)
            
        with gr.Column():
            gr.Markdown("""
            ### Named Entity Recognition Demo
            Using [token_classification/distilbert](https://sparsezoo.neuralmagic.com/models/nlp%2Ftoken_classification%2Fdistilbert-none%2Fpytorch%2Fhuggingface%2Fconll2003%2Fpruned80_quant-none-vnni)

            """)
            text = gr.Text(label="Text")
            btn = gr.Button("Submit")
        
            sparse_answers = gr.HighlightedText(label="Sparse model answers")
            sparse_duration = gr.Number(label="Sparse Latency (ms):")
            gr.Examples( [["We are flying from Texas to California"],["Mary is flying from Nairobi to New York"],["Norway is beautiful and has great hotels"] ],inputs=[text],)

            

    
    btn.click(
        run_pipeline,
        inputs=[text],
        outputs=[sparse_answers,sparse_duration],
    )

if __name__ == "__main__":
    demo.launch()