Spaces:
Sleeping
Sleeping
linh-truong
commited on
Commit
•
e276af2
1
Parent(s):
506c323
init
Browse files- .gitignore +2 -0
- app.py +51 -0
- requirements.txt +2 -0
- src/model.py +28 -0
.gitignore
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
*test*
|
2 |
+
__pycache__
|
app.py
ADDED
@@ -0,0 +1,51 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
|
3 |
+
mapper = {
|
4 |
+
"wikilingua": "64b53c9e04b2bfaeb2f0b38cb7712bcbd4755c3f",
|
5 |
+
"vietnews": "d8d516ad2c112b429155c3e6077182ae5fe5b33d"
|
6 |
+
}
|
7 |
+
|
8 |
+
if "model" not in st.session_state:
|
9 |
+
from src.model import Model
|
10 |
+
st.session_state.task="wikilingua"
|
11 |
+
model = Model(revision=mapper["wikilingua"])
|
12 |
+
st.session_state.model = model
|
13 |
+
|
14 |
+
|
15 |
+
st.set_page_config(page_title="ViT5 Reproduce", layout="wide")
|
16 |
+
hide_menu_style = """
|
17 |
+
<style>
|
18 |
+
footer {visibility: hidden;}
|
19 |
+
</style>
|
20 |
+
"""
|
21 |
+
st.markdown(hide_menu_style, unsafe_allow_html= True)
|
22 |
+
|
23 |
+
with st.sidebar:
|
24 |
+
task = st.selectbox(label="Task", options=["wikilingua", "vietnews"])
|
25 |
+
if task != st.session_state.task:
|
26 |
+
from src.model import Model
|
27 |
+
st.session_state.task=task
|
28 |
+
model = Model(revision=mapper[task])
|
29 |
+
|
30 |
+
left, middle, right = st.columns([4,1,4])
|
31 |
+
|
32 |
+
left_container = left.container(border=True)
|
33 |
+
left_container.write(f"**Input**")
|
34 |
+
left_container.divider()
|
35 |
+
text = left_container.text_area(label="", height=512, label_visibility="hidden", max_chars=4096*5)
|
36 |
+
|
37 |
+
summary_button = middle.button("Summary ➩", type="primary", use_container_width=True)
|
38 |
+
|
39 |
+
right_container = right.container(border=True)
|
40 |
+
right_container.markdown(f"**Output**")
|
41 |
+
right_container.divider()
|
42 |
+
|
43 |
+
if summary_button:
|
44 |
+
|
45 |
+
output = st.session_state.model.inference(text=text)
|
46 |
+
|
47 |
+
|
48 |
+
st.session_state["output"] = output
|
49 |
+
|
50 |
+
if "output" in st.session_state:
|
51 |
+
right_container.text_area(label="", value=st.session_state["output"], height=512, label_visibility="hidden")
|
requirements.txt
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
streamlit==1.35.0
|
2 |
+
transformers==4.41.0
|
src/model.py
ADDED
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
3 |
+
|
4 |
+
class Model:
|
5 |
+
def __init__(self, revision) -> None:
|
6 |
+
|
7 |
+
self.tokenizer = AutoTokenizer.from_pretrained("VietAI/vit5-base")
|
8 |
+
self.model = AutoModelForSeq2SeqLM.from_pretrained("truong-xuan-linh/vit5-reproduce", revision=revision)
|
9 |
+
|
10 |
+
def preprocess_function(self, text):
|
11 |
+
inputs = self.tokenizer(
|
12 |
+
text, max_length=1024, truncation=True, padding=True, return_tensors="pt"
|
13 |
+
)
|
14 |
+
return inputs
|
15 |
+
|
16 |
+
def inference(self, text):
|
17 |
+
max_target_length = 256
|
18 |
+
inputs = self.preprocess_function(text)
|
19 |
+
outputs = self.model.generate(
|
20 |
+
input_ids=inputs['input_ids'],
|
21 |
+
max_length=max_target_length,
|
22 |
+
attention_mask=inputs['attention_mask'],
|
23 |
+
)
|
24 |
+
|
25 |
+
with self.tokenizer.as_target_tokenizer():
|
26 |
+
outputs = [self.tokenizer.decode(out, clean_up_tokenization_spaces=False, skip_special_tokens=True) for out in outputs]
|
27 |
+
|
28 |
+
return outputs[0]
|