File size: 1,660 Bytes
6631f7e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9499801
6631f7e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
from transformers import GPT2LMHeadModel, GPT2Tokenizer
import torch
import streamlit as st

model = GPT2LMHeadModel.from_pretrained(
    'sberbank-ai/rugpt3small_based_on_gpt2',
    output_attentions = False,
    output_hidden_states = False,
)

tokenizer = GPT2Tokenizer.from_pretrained('sberbank-ai/rugpt3small_based_on_gpt2')

# Вешаем сохраненные веса на нашу модель
model.load_state_dict(torch.load('model.pt', map_location=torch.device('cpu')))

prompt = st.text_input('Введите текст prompt:')
length = st.slider('Длина генерируемой последовательности:', 1, 256, 16)
num_samples = st.slider('Число генераций:', 1, 6, 1)
temperature = st.slider('Температура:', 1.0, 6.0, 1.0)
selected_text = st.empty()

def generate_text(model, tokenizer, prompt, length, num_samples, temperature, selected_text):
    input_ids = tokenizer.encode(prompt, return_tensors='pt')
    output_sequences = model.generate(
        input_ids=input_ids,
        max_length=length,
        num_return_sequences=num_samples,
        temperature=temperature
    )

    generated_texts = []
    for output_sequence in output_sequences:
        generated_text = tokenizer.decode(output_sequence, clean_up_tokenization_spaces=True)
        generated_texts.append(generated_text)

    selected_text.slider('Выберите текст:', 1, num_samples, 1)
    return generated_texts[selected_text.value-1]

if st.button('Сгенерировать текст'):
    text = generate_text(model, tokenizer, prompt, length, num_samples, temperature, selected_text)
    st.write(text)