Update README.md

README.md

@@ -1,4 +1,57 @@
 ---
 license: mit
 pipeline_tag: token-classification
----
+language:
+- en
+---\
+
+
+## Usage
+
+```python
+
+from transformers import AutoModelForTokenClassification, AutoTokenizer
+import torch
+
+
+model = AutoModelForTokenClassification.from_pretrained('Sinanmz/toxicity_token_classifier')
+tokenizer = AutoTokenizer.from_pretrained('Sinanmz/toxicity_token_classifier')
+
+def test_model(text):
+    inputs = tokenizer(text, return_tensors='pt')
+    with torch.no_grad():
+        outputs = model(**inputs)
+    logits = outputs.logits
+    predictions = np.argmax(logits.detach().numpy(), axis=2)
+    tokens = tokenizer.convert_ids_to_tokens(inputs["input_ids"][0])
+    labels = predictions[0]
+    labels = labels[1:-1]
+    tokens = tokens[1:-1]
+    result = []
+    for i in range(len(labels)):
+        if i > 0 and inputs.word_ids()[i+1] == inputs.word_ids()[i]:
+            result.popitem()
+            result.append((tokens[i-1] + tokens[i][2:], model.config.id2label[labels[i-1]]))
+        else:
+            result.append((tokens[i], model.config.id2label[labels[i]]))
+    return result
+
+
+text1 = 'Your face is disgusting.'
+print("Result:", test_model(text1))
+# output: 
+# Result: {'your': 'none', 'face': 'none', 'is': 'none', 'disgusting': 'other toxicity', '.': 'none'}
+
+
+text2 = 'What an ugly person you are.'
+print("Result:", test_model(text2))
+# output: 
+# Result: {'what': 'none', 'an': 'none', 'ugly': 'insult', 'person': 'none', 'you': 'none', 'are': 'none', '.': 'none'}
+
+
+text3 = 'Nice to meet you, sir.'
+print("Result:", test_model(text3))
+# output: 
+# Result: {'nice': 'none', 'to': 'none', 'meet': 'none', 'you': 'none', ',': 'none', 'sir': 'none', '.': 'none'}
+
+```