Update app.py

app.py

@@ -1,6 +1,7 @@
 import gradio as gr
 from transformers import MarianMTModel, MarianTokenizer
 import torch
+import re
 
 # Load the model and tokenizer from the Hub
 model_name = "Dddixyy/latin-italian-translatorV2"
@@ -9,14 +10,24 @@ model = MarianMTModel.from_pretrained(model_name)
 
 # Translation function
 def translate_latin_to_italian(latin_text):
-    # Make the first letter lowercase if the input is not empty
-    if latin_text:
-        latin_text = latin_text[0].lower() + latin_text[1:]
-    inputs = tokenizer(latin_text, return_tensors="pt", padding=True, truncation=True)
-    with torch.no_grad():
-        generated_ids = model.generate(inputs["input_ids"])
-    translation = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)
-    return translation[0]
+    # Split input text into sentences while preserving line breaks
+    sentences = re.split(r'(?<=[.!?]) +', latin_text.strip())
+    
+    translated_sentences = []
+    
+    for sentence in sentences:
+        # Make the first letter lowercase if the sentence is not empty
+        if sentence:
+            sentence = sentence[0].lower() + sentence[1:]
+        inputs = tokenizer(sentence, return_tensors="pt", padding=True, truncation=True)
+        with torch.no_grad():
+            generated_ids = model.generate(inputs["input_ids"])
+        translation = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)
+        translated_sentences.append(translation[0])
+    
+    # Reassemble the translated sentences and keep original line breaks
+    translated_text = ' '.join(translated_sentences)
+    return translated_text
 
 # Define the Gradio interface
 interface = gr.Interface(
@@ -35,4 +46,4 @@ interface = gr.Interface(
 
 # Launch the app
 if __name__ == "__main__":
-    interface.launch()
+    interface.launch()