Update app.py

app.py

@@ -18,22 +18,19 @@ def get_model():
     return model
 
 
-default_text = '한국어는 저한테 너무 어려운 언어이었어요.'
-default_text_1 = '제가 정식으로 한국말 배워 안했어요.'
+default_text = '한국어는 저한테 너무 어려운 언어이었어요. 저는 한국말 배워 안헀어요.'
+
 model = get_model()
 tokenizer = tokenizer()
 st.title("Grammatical Error Correction for Korean: Demo")
 
-text = st.text_input("Input corrputed sentence 1 :", value=default_text)
-text1 = st.text_input("Input corrputed sentence 2 :", value=default_text_1)
+text = st.text_input("Input corrputed sentence :", value=default_text)
 
-st.markdown("## Original sentence 1:")
+st.markdown("## Original sentence:")
 st.write(text)
-st.markdown("## Original sentence 2:")
-st.write(text1)
 
 if text:
-    st.markdown("## Corrected output 1")
+    st.markdown("## Corrected output")
     with st.spinner('processing..'):
         raw_input_ids = tokenizer.encode(text)
         input_ids = [tokenizer.bos_token_id] + \
@@ -42,24 +39,7 @@ if text:
                                      max_length=256,
                                      eos_token_id=1,
                                      num_beams=4,
+                                     early_stopping=True,
                                      repetition_penalty=2.0)
-        output = tokenizer.decode(corrected_ids.squeeze().tolist(), skip_special_tokens=True)
-        if output == '':
-            output = 'Nothing generated...TT Please try again with different text!'
-    st.write(output)
-
-if text1:
-    st.markdown("## Corrected output 2")
-    with st.spinner('processing..'):
-        raw_input_ids1 = tokenizer.encode(text)
-        input_ids1 = [tokenizer.bos_token_id] + \
-            raw_input_ids1 + [tokenizer.eos_token_id]
-        corrected_ids1 = model.generate(torch.tensor([input_ids1]),
-                                     max_length=256,
-                                     eos_token_id=1,
-                                     num_beams=4,
-                                     repetition_penalty=2.0)
-        output1 = tokenizer.decode(corrected_ids1.squeeze().tolist(), skip_special_tokens=True)
-        if output1 == '':
-            output1 = 'Nothing generated...TT Please try again with different text!'
-    st.write(output1)
+        summ = tokenizer.decode(corrected_ids.squeeze().tolist(), skip_special_tokens=True)
+    st.write(summ)