feat: Add inference code for the Topic Classifier model

Added `model_fn` and `predict_fn` functions to load the model and run inference. Updated `README.md` to include the new inference instructions and usage example.

README.md

@@ -129,6 +129,36 @@ The model's evaluation results are as follows:
 - **Evaluation Samples Per Second:** 151.586
 - **Evaluation Steps Per Second:** 2.391
 
+#### Inference Code
+
+```python
+from transformers import AutoModelForSequenceClassification, AutoTokenizer, pipeline
+
+
+def model_fn(model_dir):
+    """
+    Load the model and tokenizer from the specified paths
+    :param model_dir:
+    :return:
+    """
+    tokenizer = AutoTokenizer.from_pretrained(model_dir)
+    model = AutoModelForSequenceClassification.from_pretrained(model_dir)
+    return model, tokenizer
+
+
+def predict_fn(data, model_and_tokenizer):
+    # destruct model and tokenizer
+    model, tokenizer = model_and_tokenizer
+
+    bert_pipe = pipeline("text-classification", model=model, tokenizer=tokenizer,
+                         truncation=True, max_length=512, return_all_scores=True)
+    # Tokenize the input, pick up first 512 tokens before passing it further
+    tokens = tokenizer.encode(data['inputs'], add_special_tokens=False, max_length=512, truncation=True)
+    input_data = tokenizer.decode(tokens)
+    return bert_pipe(input_data)
+
+```
+
 ## Conclusion
 
 The Topic Classifier achieves high accuracy, precision, recall, and F1-score, making it a reliable model for categorizing text across the domains of corporate documents, financial content, harmful content, and medical texts. The model is optimized for immediate deployment and works efficiently in real-world applications.

code/code_inference.py

@@ -0,0 +1,24 @@
+from transformers import AutoModelForSequenceClassification, AutoTokenizer, pipeline
+
+
+def model_fn(model_dir):
+    """
+    Load the model and tokenizer from the specified paths
+    :param model_dir:
+    :return:
+    """
+    tokenizer = AutoTokenizer.from_pretrained(model_dir)
+    model = AutoModelForSequenceClassification.from_pretrained(model_dir)
+    return model, tokenizer
+
+
+def predict_fn(data, model_and_tokenizer):
+    # destruct model and tokenizer
+    model, tokenizer = model_and_tokenizer
+
+    bert_pipe = pipeline("text-classification", model=model, tokenizer=tokenizer,
+                         truncation=True, max_length=512, return_all_scores=True)
+    # Tokenize the input, pick up first 512 tokens before passing it further
+    tokens = tokenizer.encode(data['inputs'], add_special_tokens=False, max_length=512, truncation=True)
+    input_data = tokenizer.decode(tokens)
+    return bert_pipe(input_data)