Update fn.py

fn.py

@@ -9,14 +9,10 @@ torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
 device = "cuda:0" if torch.cuda.is_available() else "cpu"
 
 model = None
-model_size = None
 pipe = None
 
-def load_model(_model_size):
-    global model_size, model, pipe
-
-    if _model_size and model_size != _model_size:
-        model_size = _model_size
+def load_model():
+    global model, pipe
 
     # load model
     model = AutoModelForSpeechSeq2Seq.from_pretrained(model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True)
@@ -33,9 +29,10 @@ def load_model(_model_size):
     )
 
 def speech_to_text(audio_file, _model_size = None):
-    global model_size, model, pipe
+    global model, pipe
 
-    load_model(_model_size)
+    if not model:
+        load_model()
 
     # run inference
     result = pipe(audio_file)