Spaces:
Runtime error
Runtime error
Commit
·
12b0f88
1
Parent(s):
88c8380
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,44 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
|
3 |
+
#for reproducability
|
4 |
+
SEED = 12
|
5 |
+
|
6 |
+
#maximum number of words in output text
|
7 |
+
# MAX_LEN = 30
|
8 |
+
|
9 |
+
title = st.text_input('Enter the seed words', ' ')
|
10 |
+
input_sequence = title
|
11 |
+
|
12 |
+
number = st.number_input('Insert how many words', 1)
|
13 |
+
MAX_LEN = number
|
14 |
+
|
15 |
+
if st.button('Submit'):
|
16 |
+
|
17 |
+
#get transformers
|
18 |
+
from transformers import TFGPT2LMHeadModel, GPT2Tokenizer
|
19 |
+
|
20 |
+
tokenizer = GPT2Tokenizer.from_pretrained("gpt2-medium")
|
21 |
+
GPT2 = TFGPT2LMHeadModel.from_pretrained("gpt2-medium", pad_token_id=tokenizer.eos_token_id)
|
22 |
+
|
23 |
+
import tensorflow as tf
|
24 |
+
tf.random.set_seed(SEED)
|
25 |
+
input_ids = tokenizer.encode(input_sequence, return_tensors='tf')
|
26 |
+
|
27 |
+
#sample only from 80% most likely words
|
28 |
+
sample_output = GPT2.generate(
|
29 |
+
input_ids,
|
30 |
+
do_sample = True,
|
31 |
+
max_length = MAX_LEN,
|
32 |
+
top_p = 0.8,
|
33 |
+
top_k = 0
|
34 |
+
)
|
35 |
+
|
36 |
+
st.write(tokenizer.decode(sample_output[0], skip_special_tokens = True))
|
37 |
+
|
38 |
+
else:
|
39 |
+
st.write(' ')
|
40 |
+
|
41 |
+
|
42 |
+
|
43 |
+
# print("Output:\n" + 100 * '-')
|
44 |
+
# print(tokenizer.decode(sample_output[0], skip_special_tokens = True), '...')
|