tskolm commited on
Commit
fd88ce0
·
1 Parent(s): 0831deb

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +13 -8
app.py CHANGED
@@ -9,15 +9,20 @@ from transformers import GPT2Tokenizer, GPT2LMHeadModel, GPT2Config
9
 
10
  def generate(tokenizer, model, text, features):
11
  generated = tokenizer("<|startoftext|><|titlestart|>{}<|titleend|><|authornamebegin|>".format(text), return_tensors="pt").input_ids
12
- sample_outputs = model.generate(
13
- generated, do_sample=True, top_k=50,
14
- max_length=features['max_length'], top_p=features['top_p'], temperature=features['t'] / 100.0, num_return_sequences=features['num'],
15
- )
16
- for i, sample_output in enumerate(sample_outputs):
17
- decoded = tokenizer.decode(sample_output, skip_special_tokens=False)
18
- if '<|authornameend|>' not in decoded:
 
 
19
  continue
20
- author, text = decoded.split('<|authornamebegin|>')[1].split('<|authornameend|>')
 
 
 
21
  st.markdown('**' + author.strip() + '**: ' + text.replace('<|endoftext|>', '').replace('<|pad|>', '').strip())
22
 
23
 
 
9
 
10
  def generate(tokenizer, model, text, features):
11
  generated = tokenizer("<|startoftext|><|titlestart|>{}<|titleend|><|authornamebegin|>".format(text), return_tensors="pt").input_ids
12
+ count = 0
13
+ while count < features['num']:
14
+ sample_outputs = model.generate(
15
+ generated, do_sample=True, top_k=50,
16
+ max_length=features['max_length'], top_p=features['top_p'], temperature=features['t'] / 100.0, num_return_sequences=1,
17
+ )
18
+ decoded = tokenizer.decode(sample_outputs[0], skip_special_tokens=False)
19
+ print(decoded, file=sys.stderr)
20
+ if '<|authornameend|>' not in decoded:
21
  continue
22
+
23
+ author, text = decoded.split('<|authornamebegin|>')[-1].split('<|authornameend|>')
24
+
25
+ count += 1
26
  st.markdown('**' + author.strip() + '**: ' + text.replace('<|endoftext|>', '').replace('<|pad|>', '').strip())
27
 
28