Update app.py

app.py

@@ -1,14 +1,10 @@
 import gradio as gr
-from flores import code_mapping
 from functools import lru_cache
 import openai  # 用于调用外部API
 import os
 import spaces
 import gradio as gr
-from sacremoses import MosesPunctNormalizer
-from stopes.pipelines.monolingual.utils.sentence_split import get_split_algo
 from transformers import AutoTokenizer, AutoModel,AutoModelForCausalLM
-from flores import code_mapping
 import platform
 import torch
 import nltk
@@ -51,37 +47,21 @@ def translate(text: str, src_lang: str, tgt_lang: str):
 # Only assign GPU if cache not used
 @spaces.GPU
 def _translate(text: str, src_lang: str, tgt_lang: str):
-    paragraphs = text.split("\n")
-    translated_paragraphs = []
-
-    for paragraph in paragraphs:
-        translated_sentences = []
-        input_tokens = (
-            tokenizer("Translate to Chinese. Direct output translation result without any explaination::\n\n" + paragraph, return_tensors="pt")
-            .input_ids[0]
-            .cpu()
-            .numpy()
-            .tolist()
-        )
-        translated_chunk = model.generate(
-            input_ids=torch.tensor([input_tokens]).to(device),
-            max_length=len(input_tokens) + 1000,
-            num_return_sequences=1,
-        )
-        print(translated_chunk)
-        translated_chunk = tokenizer.batch_decode(
-            translated_chunk[0], skip_special_tokens=True
-        )
-        
-        if isinstance(translated_chunk,list):
-            translated_chunk = "".join(translated_chunk)
-        translated_sentences.append(translated_chunk)
-        print("dev: ",translated_chunk)
-        translated_paragraph = " ".join(translated_sentences)
-        translated_paragraphs.append(translated_paragraph)
-
-    return "\n".join(translated_paragraphs)
-
+    input_tokens = (
+        tokenizer(text, return_tensors="pt")
+        .input_ids[0]
+        .cpu()
+        .numpy()
+        .tolist()
+    )
+    translated_chunk = model.generate(
+        input_ids=torch.tensor([input_tokens]).to(device),
+        max_length=len(input_tokens) + 1000,
+        num_return_sequences=1,
+    )
+    full_output = tokenizer.decode(translated_chunk[0], skip_special_tokens=True).strip()
+    print(full_output)
+    return full_output
 
 # def _translate(text: str, src_lang: str, tgt_lang: str):
 #     prompt = f"Translate the following text from {src_lang} to {tgt_lang}. Direct output translation result without any explaination:\n\n{text}"