Update app.py

app.py

@@ -50,6 +50,17 @@ def generate(model,tokenizer,instruction,context,streamer):
             output = output.split("###")[3].split("<|endoftext|>", 1)[0]
             return output
 
+
+@st.experimental_singleton
+def get_model():
+    # Create a model of the specified type.
+    return AutoModelForCausalLM.from_pretrained("lagy/carballo_finetuned")
+
+@st.experimental_singleton
+def get_tokenizer():
+    # Create a model of the specified type.
+    return AutoTokenizer.from_pretrained("lagy/carballo_finetuned")
+
 st.write(" ")
 
 instruction = st.text_area('Instrucción')
@@ -57,8 +68,8 @@ context = st.text_area('Contexto')
 max_length = st.number_input('Max generation length',value=10)
 
 
-model = AutoModelForCausalLM.from_pretrained("lagy/carballo_finetuned")
-tokenizer = AutoTokenizer.from_pretrained("lagy/carballo_finetuned")
+model = get_model()
+tokenizer = get_tokenizer()
 model.eval()