add usage example

README.md

@@ -15,11 +15,34 @@ widget:
 
 Repository: [https://github.com/Lurunchik/NF-CATS](https://github.com/Lurunchik/NF-CATS)
 
-
 Model trained with NFQA dataset. Base model is [roberta-base-squad2](https://huggingface.co/deepset/roberta-base-squad2), a RoBERTa-based model for the task of Question Answering, fine-tuned using the SQuAD2.0 dataset.
 
 Uses `NOT-A-QUESTION`, `FACTOID`, `DEBATE`, `EVIDENCE-BASED`, `INSTRUCTION`, `REASON`, `EXPERIENCE`, `COMPARISON` labels.
 
+## How to use NFQA cat with HuggingFace
+
+##### Load NFQA cat and its tokenizer:
+```python
+from transformers import AutoTokenizer
+
+from nfqa_model import RobertaNFQAClassification 
+
+nfqa_model = RobertaNFQAClassification.from_pretrained("Lurunchik/nf-cats")
+nfqa_tokenizer = AutoTokenizer.from_pretrained("deepset/roberta-base-squad2")
+```
+
+##### Make prediction using helper function
+```python
+
+def get_nfqa_category_prediction(text):
+    output = nfqa_model(**nfqa_tokenizer(text, return_tensors="pt"))
+    index = output.logits.argmax()
+    return nfqa_model.config.id2label[int(index)]
+
+get_nfqa_category_prediction('how to assign category?')
+# result
+#'INSTRUCTION'
+```
 
 ## Citation