Update app.py

app.py

@@ -558,7 +558,7 @@ def summarize(
     text, max_len=20, min_len=10
 ):
     log(f'CALL summarize')
-    inputs = tokenizer.encode("summarize: " + text, return_tensors="pt", max_length=float('inf'), truncation=False).input_ids
+    inputs = tokenizer.encode("summarize: " + text, return_tensors="pt", max_length=float('inf'), truncation=False)
     if get_tensor_length(inputs) < 3:
         print("Summarization Error: Text is too short, 3 words minimum!")
         return text
@@ -640,15 +640,14 @@ def translate(txt,to_lang="en",from_lang=False):
     if(from_lang == to_lang):
         log(f'RET translate with txt as {txt}')
         return txt
-    inputs = tokenizer.encode(f"translate {from_lang} to {to_lang}: " + txt, return_tensors="pt", max_length=float('inf'), truncation=False).input_ids
+    inputs = tokenizer.encode(f"translate {from_lang} to {to_lang}: " + txt, return_tensors="pt", max_length=float('inf'), truncation=False)
     chunks_length = math.ceil(get_tensor_length(inputs) / 512)
     ret = ""
     for index in range(chunks_length):
-        ret = ret + ("" if ret == "" else " ") + tokenizer.decode(
-            model.generate(
-                torch.tensor([list(inputs[0][ index*512:index*512+512 ])])
-            )[0], skip_special_tokens=True
-        )
+        chunk = torch.tensor([list(inputs[0][ index*512:index*512+512 ])])
+        gen = model.generate(chunk)
+        toks = tokenizer.decode(gen[0], skip_special_tokens=True)
+        ret = ret + ("" if ret == "" else " ") + toks
     log(f'RET translate with ret as {ret}')
     return ret