Update app.py

app.py

@@ -1,8 +1,8 @@
-# Turkish Zero-Shot Text Classification with XLM-RoBERTa
+# Zero-Shot Text Classification with Multilingual T5 (mT5)
 
-from transformers import AutoTokenizer, AutoModelForSequenceClassification, pipeline
+from torch.nn.functional import softmax
+from transformers import MT5ForConditionalGeneration, MT5Tokenizer
 import sentencepiece
-import torch
 import plotly.graph_objects as go
 import streamlit as st
 
@@ -28,20 +28,13 @@ label_list_2 = ["positive", "negative", "neutral"]
 st.title("Turkish Zero-Shot Text Classification \
     with Multilingual XLM-RoBERTa and mDeBERTa Models")
 
-model_list = ['vicgalle/xlm-roberta-large-xnli-anli',
-             'joeddav/xlm-roberta-large-xnli',
-             'MoritzLaurer/mDeBERTa-v3-base-xnli-multilingual-nli-2mil7']
+model_name = "alan-turing-institute/mt5-large-finetuned-mnli-xtreme-xnli"
 
-st.sidebar.header("Select Model")
-model_checkpoint = st.sidebar.radio("", model_list)
+st.sidebar.write("For details of used model:")
+st.sidebar.write("https://huggingface.co/alan-turing-institute/mt5-large-finetuned-mnli-xtreme-xnli")
 
-st.sidebar.write("For details of models:")
-st.sidebar.write("https://huggingface.co/vicgalle")
-st.sidebar.write("https://huggingface.co/joeddav")
-st.sidebar.write("https://huggingface.co/MoritzLaurer")
-
-st.sidebar.write("For XNLI Dataset:")
-st.sidebar.write("https://huggingface.co/datasets/xnli")
+st.sidebar.write("For Xtreme XNLI Dataset:")
+st.sidebar.write("https://www.tensorflow.org/datasets/catalog/xtreme_xnli")
 
 st.subheader("Select Text and Label List")
 st.text_area("Text #1", text_1, height=128)
@@ -63,9 +56,10 @@ elif labels == "New Label List":
     selected_labels = st.text_area("New Label List (Pls Input as comma-separated)", value="", height=16).split(",")
 
 @st.cache(allow_output_mutation=True)
-def setModel(model_checkpoint):
-    model = AutoModelForSequenceClassification.from_pretrained(model_checkpoint)
-    tokenizer = AutoTokenizer.from_pretrained(model_checkpoint)
+def setModel(model_name):
+    tokenizer = MT5Tokenizer.from_pretrained(model_name)
+    model = MT5ForConditionalGeneration.from_pretrained(model_name)
+    model.eval()
     return pipeline("zero-shot-classification", model=model, tokenizer=tokenizer)
         
 Run_Button = st.button("Run", key=None)