Create handler.py

handler.py

@@ -0,0 +1,16 @@
+model_name_or_path="alfaxadeyembe/gemma2-2b-swahili-it"
+from transformers import AutoModelForSequenceClassification, AutoTokenizer
+import torch
+
+class EndpointHandler:
+    def __init__(self, model_name_or_path):
+        self.tokenizer = AutoTokenizer.from_pretrained(model_name_or_path)
+        self.model = AutoModelForSequenceClassification.from_pretrained(model_name_or_path)
+        self.model.eval()  # Set the model to evaluation mode
+
+    def __call__(self, data):
+        inputs = data.get("inputs", "")
+        tokens = self.tokenizer(inputs, return_tensors='pt')
+        with torch.no_grad():
+            outputs = self.model(**tokens)
+        return outputs