create models.py with AlbertQA model

app/models.py

@@ -0,0 +1,66 @@
+from transformers import AutoTokenizer, AlbertForQuestionAnswering
+import torch
+
+class QAModel:
+    def __init__(self, model_name: str = "twmkn9/albert-base-v2-squad2"):
+        """
+        Initialize the QA model and tokenizer.
+        """
+        self.model_name = model_name
+        self.tokenizer = None
+        self.model = None
+        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+        self.load_model()
+
+    def load_model(self):
+        """
+        Load the tokenizer and model.
+        """
+        self.tokenizer = AutoTokenizer.from_pretrained(self.model_name)
+        self.model = AlbertForQuestionAnswering.from_pretrained(self.model_name).to(self.device)
+        logger.info(f"Loaded QA model: {self.model_name}")
+
+    def inference_qa(self, context: str, question: str):
+        """
+        Perform question-answering inference.
+        Args:
+            context (str): The text passage or document.
+            question (str): The question to be answered.
+        Returns:
+            str: The predicted answer.
+        """
+        if self.tokenizer is None or self.model is None:
+            raise ValueError("Model or tokenizer is not loaded.")
+
+        # Tokenize inputs
+        inputs = self.tokenizer(question, context, return_tensors="pt", truncation=True, padding=True)
+        inputs = {key: value.to(self.device) for key, value in inputs.items()}
+
+        # Perform inference
+        with torch.no_grad():
+            outputs = self.model(**inputs)
+
+        # Extract answer
+        answer_start_index = outputs.start_logits.argmax()
+        answer_end_index = outputs.end_logits.argmax()
+        predict_answer_tokens = inputs["input_ids"][0, answer_start_index : answer_end_index + 1]
+        answer = self.tokenizer.decode(predict_answer_tokens, skip_special_tokens=True)
+
+        return answer
+
+# Global instance of the QA model
+qa_model = QAModel()
+
+def load_qa_pipeline(model_name: str = "twmkn9/albert-base-v2-squad2"):
+    """
+    Load the QA model and tokenizer.
+    """
+    global qa_model
+    qa_model = QAModel(model_name)
+    return qa_model
+
+def inference_qa(qa_pipeline, context: str, question: str):
+    """
+    Perform QA inference using the loaded pipeline.
+    """
+    return qa_pipeline.inference_qa(context, question)