Update app.py

app.py

@@ -561,7 +561,7 @@ def summarize(
     
     words = text.split()
     
-    if get_tensor_length(words) < 5:
+    if len(words) < 5:
         print("Summarization Error: Text is too short, 5 words minimum!")
         return text
         
@@ -571,7 +571,7 @@ def summarize(
     for index in range(math.ceil( len(words) / 512 )):
         
         chunk = " ".join(words[ index*512:(index+1)*512 ])
-        inputs = tokenizer.encode( prefix + chunk, return_tensors="pt", max_length=float('inf'), truncation=False)
+        inputs = tokenizer.encode( prefix + chunk, return_tensors="pt", truncation=False)
         
         while get_tensor_length(inputs) > max_len:
             
@@ -587,7 +587,7 @@ def summarize(
         toks = tokenizer.decode(inputs[0], skip_special_tokens=True)
         ret = ret + ("" if ret == "" else " ") + toks
 
-    inputs = tokenizer.encode( prefix + ret, return_tensors="pt", max_length=float('inf'), truncation=False)
+    inputs = tokenizer.encode( prefix + ret, return_tensors="pt", truncation=False)
     gen = model.generate(
         inputs,
         length_penalty=1.0,
@@ -723,9 +723,10 @@ def translate(txt,to_lang="en",from_lang=False):
     words = txt.split()
     ret = ""
     for index in range(math.ceil( len(words) / 512 )):
-        chunk = " ".join(words[ index*512:(index+1)*512 ])
-        inputs = tokenizer.encode(prefix+chunk, return_tensors="pt", max_length=float('inf'), truncation=False)
-        gen = model.generate(inputs)
+        chunk = " ".join(words[index*512:(index+1)*512])
+        log(f'DBG translate chunk is {chunk}')
+        inputs = tokenizer.encode(prefix+chunk, return_tensors="pt", truncation=False)
+        gen = model.generate(inputs,num_beams=3)
         toks = tokenizer.decode(gen[0], skip_special_tokens=True)
         ret = ret + ("" if ret == "" else " ") + toks
     log(f'RET translate with ret as {ret}')