Refactor translate function to split input text into smaller chunks

app.py

@@ -21,7 +21,7 @@ model, tokenizer = load_model()
 
 
 @spaces.GPU
-def translate(text: str, src_lang: str, tgt_lang: str):
+def _translate(text: str, src_lang: str, tgt_lang: str):
     source = code_mapping[src_lang]
     target = code_mapping[tgt_lang]
     translator = pipeline(
@@ -36,6 +36,18 @@ def translate(text: str, src_lang: str, tgt_lang: str):
     return output[0]["translation_text"]
 
 
+def translate(text: str, src_lang: str, tgt_lang: str):
+    # split the input text into smaller chunks
+    # split first on newlines
+    outputs = ""
+    chunks = text.split("\n")
+    for chunk in chunks:
+        # run the translation on each chunk
+        output = _translate(chunk, src_lang, tgt_lang)
+        outputs += output + "\n"
+    return outputs
+
+
 description = """
 No Language Left Behind (NLLB) is a series of open-source models aiming to provide high-quality translations between 200 language."""