Upload handler.py

handler.py

@@ -0,0 +1,22 @@
+import torch
+from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
+
+class ModelHandler:
+    def __init__(self):
+        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+        self.model = AutoModelForSeq2SeqLM.from_pretrained("shaheerzk/text_to_sql")
+        self.tokenizer = AutoTokenizer.from_pretrained("shaheerzk/text_to_sql")
+        self.model.to(self.device)
+
+    def handle(self, inputs):
+        # Preprocess input
+        text = inputs.get("text", "")
+        inputs = self.tokenizer(text, return_tensors="pt").to(self.device)
+
+        # Inference
+        with torch.no_grad():
+            outputs = self.model.generate(**inputs)
+
+        # Post-process output
+        generated_text = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
+        return {"generated_text": generated_text}