Update app.py

app.py

@@ -92,9 +92,14 @@ def create_response_untethered(input_str,
   #                                                      num_beams=num_beams,
   #                                                      num_return_sequences=num_return_sequences)[0])
 
-  input_ids = tokenizer.encode(input_str + tokenizer.eos_token, return_tensors="pt")
+  #input_ids = tokenizer.encode(input_str + tokenizer.eos_token, return_tensors="pt")
+  encoded = tokenizer.encode_plus(input_str + tokenizer.eos_token, return_tensors="pt")
+  input_ids = encoded["input_ids"]
+  attention_mask = encoded["attention_mask"]
+
+
   #output_ids = fine_tuned_model.generate(input_ids,do_sample=True, max_length=100, temperature=0.2, top_p=0.9, repetition_penalty=1.5,num_return_sequences=6)                                
-  output_ids = untethered_model.generate(input_ids,do_sample=do_sample, max_length=100, temperature=temperature, top_p=top_p, top_k=top_k, repetition_penalty=repetition_penalty,num_return_sequences=num_return_sequences, num_beams = num_beams)                                
+  output_ids = untethered_model.generate(input_ids,do_sample=do_sample, attention_mask=attention_mask, max_length=100, temperature=temperature, top_p=top_p, top_k=top_k, repetition_penalty=repetition_penalty,num_return_sequences=num_return_sequences, num_beams = num_beams)                                
   outputs = ""                                
   for output_id in output_ids:
     output = tokenizer.decode(output_id, skip_special_tokens=True)