load the models before the endpoint is called

main.py

@@ -22,6 +22,12 @@ app.add_middleware(
 	allow_headers=["*"]
 )
 
+tokenizer = T5Tokenizer.from_pretrained(
+			"Rostlab/prot_t5_xl_half_uniref50-enc", do_lower_case=False, torch_dtype=torch.float16)
+
+model = T5EncoderModel.from_pretrained(
+			"Rostlab/prot_t5_xl_half_uniref50-enc")
+
 @app.get("/predict")
 def predict_UniKP_values(
 	sequence: str,
@@ -41,11 +47,8 @@ def predict_UniKP_values(
 class EndpointHandler():
 	def __init__(self, path=""):
 
-		# load tokenizer and model
-		self.tokenizer = T5Tokenizer.from_pretrained(
-			"Rostlab/prot_t5_xl_half_uniref50-enc", do_lower_case=False, torch_dtype=torch.float16)
-		self.model = T5EncoderModel.from_pretrained(
-			"Rostlab/prot_t5_xl_half_uniref50-enc")
+		self.tokenizer = tokenizer
+		self.model = model
 		
 		# path to the vocab_content and trfm model
 		vocab_content_path = f"{path}/vocab_content.txt"