Update app.py

app.py

@@ -2,23 +2,30 @@ import streamlit as st
 import transformers
 import tensorflow
 
-from transformers import pipeline
 
-# Replace this with your own checkpoint
-model_checkpoint = "Modfiededition/t5-base-fine-tuned-on-jfleg"
+from transformers import AutoTokenizer
+from transformers import TFAutoModelForSeq2SeqLM
 
+model_checkpoint = "Modfiededition/t5-base-fine-tuned-on-jfleg"
+tokenizer = AutoTokenizer.from_pretrained(model_checkpoint)
 
+@st.cache
 def load_model(model_name):
-    translator= pipeline("text2text-generation", model=model_checkpoint)
-    return translator
+    model = TFAutoModelForSeq2SeqLM.from_pretrained(model_name)
+    return model
     
-translator= load_model(model_checkpoint)
+model= load_model(model_checkpoint)
 
 default_value = "Write your text here!"
 #prompts
 st.title("Writing Assistant for you 🦄")
 
 sent = st.text_area("Text", default_value, height = 275)
-generated_sequences = translator(sent)
 
-st.write(generated_sequences[-1])
+inputs = tokenizer("Grammar: "+sent,return_tensors="tf")
+
+output_ids = model.generate(inputs["input_ids"]).numpy()[0][1:-1]
+
+generated_sequences = tokenizer.decode(output_ids)
+
+st.write(generated_sequences)