vinayakdev commited on
Commit
b306dde
1 Parent(s): fa7ee8f

Update generator.py

Browse files
Files changed (1) hide show
  1. generator.py +12 -8
generator.py CHANGED
@@ -46,19 +46,23 @@ hfmodel, hftokenizer, tok, model = load_model()
46
  nlp = pipeline("e2e-qg")
47
  def run_model(input_string, **generator_args):
48
  generator_args = {
49
- "max_length": 256,
50
  "num_beams": 4,
51
- # "length_penalty": 1.5,
52
  "no_repeat_ngram_size": 2,
53
  "early_stopping": False,
54
  }
55
  # tokenizer = att.from_pretrained("ThomasSimonini/t5-end2end-question-generation")
56
- output = nlp(input_string)
57
- # input_string = "generate questions: " + input_string + " </s>"
58
- # input_ids = hftokenizer.encode(input_string, return_tensors="pt")
59
- # res = hfmodel.generate(input_ids, **generator_args)
60
- # output = hftokenizer.batch_decode(res, skip_special_tokens=True)
61
- # output = [item.split("<sep>") for item in output]
 
 
 
 
62
  return output
63
 
64
 
 
46
  nlp = pipeline("e2e-qg")
47
  def run_model(input_string, **generator_args):
48
  generator_args = {
49
+ "max_model_length": 256,
50
  "num_beams": 4,
51
+ "length_penalty": 1.5,
52
  "no_repeat_ngram_size": 2,
53
  "early_stopping": False,
54
  }
55
  # tokenizer = att.from_pretrained("ThomasSimonini/t5-end2end-question-generation")
56
+ # output = nlp(input_string)
57
+
58
+ input_string = "generate questions: " + input_string + " </s>"
59
+
60
+ inputs = hftokenizer.encode(input_string, return_tensors="pt")
61
+
62
+ res = model.generate(input_ids=inputs['input_ids'], attention_mask=inputs['attention_mask'], **generate_kwargs)
63
+ output = hftokenizer.decode(res[0], skip_special_tokens=True)
64
+ output = [item.split("<sep>") for item in output]
65
+ output = [o.strip() for o in output[:-1]]
66
  return output
67
 
68