Update pages/gpt (1).py
Browse files- pages/gpt (1).py +8 -71
pages/gpt (1).py
CHANGED
@@ -1,73 +1,10 @@
|
|
1 |
-
from transformers import GPT2LMHeadModel, GPT2Tokenizer
|
2 |
import streamlit as st
|
3 |
-
import torch
|
4 |
-
import textwrap
|
5 |
-
import plotly.express as px
|
6 |
-
|
7 |
-
from streamlit_extras.let_it_rain import rain
|
8 |
-
|
9 |
-
rain(
|
10 |
-
emoji="⭐",
|
11 |
-
font_size=54,
|
12 |
-
falling_speed=5,
|
13 |
-
animation_length="infinite",
|
14 |
-
)
|
15 |
-
|
16 |
-
st.header(':green[Text generation by GPT2 model]')
|
17 |
-
|
18 |
-
tokenizer = GPT2Tokenizer.from_pretrained('sberbank-ai/rugpt3small_based_on_gpt2')
|
19 |
-
model = GPT2LMHeadModel.from_pretrained(
|
20 |
-
'sberbank-ai/rugpt3small_based_on_gpt2',
|
21 |
-
output_attentions = False,
|
22 |
-
output_hidden_states = False,
|
23 |
-
)
|
24 |
-
|
25 |
-
model.load_state_dict(torch.load('models/model.pt', map_location=torch.device('cpu')))
|
26 |
-
|
27 |
-
|
28 |
-
length = st.sidebar.slider('**Generated sequence length:**', 8, 256, 15)
|
29 |
-
if length > 100:
|
30 |
-
st.warning("This is very hard for me, please have pity on me. Could you lower the value?", icon="🤖")
|
31 |
-
num_samples = st.sidebar.slider('**Number of generations:**', 1, 10, 1)
|
32 |
-
if num_samples > 4:
|
33 |
-
st.warning("OH MY ..., I have to work late again!!! Could you lower the value?", icon="🤖")
|
34 |
-
temperature = st.sidebar.slider('**Temperature:**', 0.1, 10.0, 3.0)
|
35 |
-
if temperature > 6.0:
|
36 |
-
st.info('What? You want to get some kind of bullshit as a result? Turn down the temperature', icon="🤖")
|
37 |
-
top_k = st.sidebar.slider('**Number of most likely generation words:**', 10, 200, 50)
|
38 |
-
top_p = st.sidebar.slider('**Minimum total probability of top words:**', 0.4, 1.0, 0.9)
|
39 |
-
|
40 |
-
|
41 |
-
prompt = st.text_input('**Enter text 👇:**')
|
42 |
-
if st.button('**Generate text**'):
|
43 |
-
image_container = st.empty()
|
44 |
-
image_container.image("pict/wait.jpeg", caption="that's so long!!!", use_column_width=True)
|
45 |
-
with torch.inference_mode():
|
46 |
-
prompt = tokenizer.encode(prompt, return_tensors='pt')
|
47 |
-
out = model.generate(
|
48 |
-
input_ids=prompt,
|
49 |
-
max_length=length,
|
50 |
-
num_beams=8,
|
51 |
-
do_sample=True,
|
52 |
-
temperature=temperature,
|
53 |
-
top_k=top_k,
|
54 |
-
top_p=top_p,
|
55 |
-
no_repeat_ngram_size=3,
|
56 |
-
num_return_sequences=num_samples,
|
57 |
-
).cpu().numpy()
|
58 |
-
image_container.empty()
|
59 |
-
st.write('**_Результат_** 👇')
|
60 |
-
for i, out_ in enumerate(out):
|
61 |
-
# audio_file = open('pict/pole-chudes-priz.mp3', 'rb')
|
62 |
-
# audio_bytes = audio_file.read()
|
63 |
-
# st.audio(audio_bytes, format='audio/mp3')
|
64 |
-
|
65 |
-
with st.expander(f'Текст {i+1}:'):
|
66 |
-
st.write(textwrap.fill(tokenizer.decode(out_), 100))
|
67 |
-
st.image("pict/wow.png")
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import streamlit as st
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
3 |
+
txt = st.text_area('Text to analyze', '''
|
4 |
+
It was the best of times, it was the worst of times, it was
|
5 |
+
the age of wisdom, it was the age of foolishness, it was
|
6 |
+
the epoch of belief, it was the epoch of incredulity, it
|
7 |
+
was the season of Light, it was the season of Darkness, it
|
8 |
+
was the spring of hope, it was the winter of despair, (...)
|
9 |
+
''')
|
10 |
+
st.write('Sentiment:', run_sentiment_analysis(txt))
|