Spaces:
Runtime error
Runtime error
vvv-knyazeva
commited on
Commit
•
6631f7e
1
Parent(s):
76e122c
Upload 2 files
Browse files- pages/gpt_v1.py +47 -0
- pages/gpt_v2.py +41 -0
pages/gpt_v1.py
ADDED
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from transformers import GPT2LMHeadModel, GPT2Tokenizer
|
2 |
+
import torch
|
3 |
+
import streamlit as st
|
4 |
+
|
5 |
+
model = GPT2LMHeadModel.from_pretrained(
|
6 |
+
'sberbank-ai/rugpt3small_based_on_gpt2',
|
7 |
+
output_attentions = False,
|
8 |
+
output_hidden_states = False,
|
9 |
+
)
|
10 |
+
|
11 |
+
tokenizer = GPT2Tokenizer.from_pretrained('sberbank-ai/rugpt3small_based_on_gpt2')
|
12 |
+
|
13 |
+
|
14 |
+
# Вешаем сохраненные веса на нашу модель
|
15 |
+
model.load_state_dict(torch.load('models/model.pt', map_location=torch.device('cpu')))
|
16 |
+
|
17 |
+
prompt = st.text_input('Введите текст prompt:')
|
18 |
+
length = st.slider('Длина генерируемой последовательности:', 10, 1000, 50)
|
19 |
+
num_samples = st.slider('Число генераций:', 1, 10, 1)
|
20 |
+
temperature = st.slider('Температура:', 0.1, 1.0, 0.5)
|
21 |
+
|
22 |
+
|
23 |
+
def generate_text(model, tokenizer, prompt, length, num_samples, temperature):
|
24 |
+
input_ids = tokenizer.encode(prompt, return_tensors='pt')
|
25 |
+
output_sequences = model.generate(
|
26 |
+
input_ids=input_ids,
|
27 |
+
max_length=length,
|
28 |
+
num_return_sequences=num_samples,
|
29 |
+
temperature=temperature
|
30 |
+
)
|
31 |
+
|
32 |
+
generated_texts = []
|
33 |
+
for output_sequence in output_sequences:
|
34 |
+
generated_text = tokenizer.decode(output_sequence, clean_up_tokenization_spaces=True)
|
35 |
+
generated_texts.append(generated_text)
|
36 |
+
|
37 |
+
return generated_texts
|
38 |
+
|
39 |
+
|
40 |
+
if st.button('Сгенерировать текст'):
|
41 |
+
generated_texts = generate_text(model, tokenizer, prompt, length, num_samples, temperature)
|
42 |
+
for i, text in enumerate(generated_texts):
|
43 |
+
st.write(f'Текст {i+1}:')
|
44 |
+
st.write(text)
|
45 |
+
|
46 |
+
|
47 |
+
|
pages/gpt_v2.py
ADDED
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from transformers import GPT2LMHeadModel, GPT2Tokenizer
|
2 |
+
import torch
|
3 |
+
import streamlit as st
|
4 |
+
|
5 |
+
model = GPT2LMHeadModel.from_pretrained(
|
6 |
+
'sberbank-ai/rugpt3small_based_on_gpt2',
|
7 |
+
output_attentions = False,
|
8 |
+
output_hidden_states = False,
|
9 |
+
)
|
10 |
+
|
11 |
+
tokenizer = GPT2Tokenizer.from_pretrained('sberbank-ai/rugpt3small_based_on_gpt2')
|
12 |
+
|
13 |
+
# Вешаем сохраненные веса на нашу модель
|
14 |
+
model.load_state_dict(torch.load('model.pt', map_location=torch.device('cpu')))
|
15 |
+
|
16 |
+
prompt = st.text_input('Введите текст prompt:')
|
17 |
+
length = st.slider('Длина генерируемой последовательности:', 10, 256, 16)
|
18 |
+
num_samples = st.slider('Число генераций:', 1, 6, 1)
|
19 |
+
temperature = st.slider('Температура:', 1.0, 6.0, 1.0)
|
20 |
+
selected_text = st.empty()
|
21 |
+
|
22 |
+
def generate_text(model, tokenizer, prompt, length, num_samples, temperature, selected_text):
|
23 |
+
input_ids = tokenizer.encode(prompt, return_tensors='pt')
|
24 |
+
output_sequences = model.generate(
|
25 |
+
input_ids=input_ids,
|
26 |
+
max_length=length,
|
27 |
+
num_return_sequences=num_samples,
|
28 |
+
temperature=temperature
|
29 |
+
)
|
30 |
+
|
31 |
+
generated_texts = []
|
32 |
+
for output_sequence in output_sequences:
|
33 |
+
generated_text = tokenizer.decode(output_sequence, clean_up_tokenization_spaces=True)
|
34 |
+
generated_texts.append(generated_text)
|
35 |
+
|
36 |
+
selected_text.slider('Выберите текст:', 1, num_samples, 1)
|
37 |
+
return generated_texts[selected_text.value-1]
|
38 |
+
|
39 |
+
if st.button('Сгенерировать текст'):
|
40 |
+
text = generate_text(model, tokenizer, prompt, length, num_samples, temperature, selected_text)
|
41 |
+
st.write(text)
|