Spaces:
Build error
Build error
EliottZemour
commited on
Commit
•
9df4338
1
Parent(s):
c88b60d
add correlation score
Browse files
app.py
CHANGED
@@ -1,23 +1,48 @@
|
|
1 |
import torch
|
2 |
import transformers
|
3 |
-
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
4 |
import gradio as gr
|
5 |
import os
|
6 |
|
7 |
model_name = 'eliolio/bart-finetuned-yelpreviews'
|
|
|
8 |
|
9 |
access_token = os.environ.get('private_token')
|
10 |
|
11 |
-
model = AutoModelForSeq2SeqLM.from_pretrained(
|
12 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
|
14 |
def create_prompt(stars, useful, funny, cool):
|
15 |
return f"Generate review: stars: {stars}, useful: {useful}, funny: {funny}, cool: {cool}"
|
16 |
|
|
|
17 |
def postprocess(review):
|
18 |
dot = review.rfind('.')
|
19 |
return review[:dot+1]
|
20 |
|
|
|
21 |
def generate_reviews(stars, useful, funny, cool):
|
22 |
text = create_prompt(stars, useful, funny, cool)
|
23 |
inputs = tokenizer(text, return_tensors='pt')
|
@@ -30,10 +55,15 @@ def generate_reviews(stars, useful, funny, cool):
|
|
30 |
top_p=0.9
|
31 |
)
|
32 |
reviews = []
|
|
|
33 |
for review in out:
|
34 |
reviews.append(postprocess(tokenizer.decode(review, skip_special_tokens=True)))
|
|
|
|
|
|
|
|
|
|
|
35 |
|
36 |
-
return reviews[0], reviews[1], reviews[2]
|
37 |
|
38 |
css = """
|
39 |
#ctr {text-align: center;}
|
@@ -65,12 +95,16 @@ demo = gr.Blocks(css=css)
|
|
65 |
with demo:
|
66 |
with gr.Row():
|
67 |
gr.Markdown(md_text)
|
68 |
-
|
69 |
with gr.Row():
|
70 |
-
stars = gr.inputs.Slider(minimum=0, maximum=5,
|
71 |
-
|
72 |
-
|
73 |
-
|
|
|
|
|
|
|
|
|
74 |
with gr.Row():
|
75 |
button = gr.Button("Generate reviews !", elem_id='btn')
|
76 |
|
@@ -79,13 +113,18 @@ with demo:
|
|
79 |
output2 = gr.Textbox(label="Review #2")
|
80 |
output3 = gr.Textbox(label="Review #3")
|
81 |
|
|
|
|
|
|
|
|
|
|
|
82 |
with gr.Row():
|
83 |
gr.Markdown(resources)
|
84 |
|
85 |
button.click(
|
86 |
fn=generate_reviews,
|
87 |
inputs=[stars, useful, funny, cool],
|
88 |
-
outputs=[output1, output2, output3]
|
89 |
)
|
90 |
|
91 |
-
demo.launch()
|
|
|
1 |
import torch
|
2 |
import transformers
|
3 |
+
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, AutoModelForSequenceClassification
|
4 |
import gradio as gr
|
5 |
import os
|
6 |
|
7 |
model_name = 'eliolio/bart-finetuned-yelpreviews'
|
8 |
+
bert_model_name = 'eliolio/bert-correlation-yelpreviews'
|
9 |
|
10 |
access_token = os.environ.get('private_token')
|
11 |
|
12 |
+
model = AutoModelForSeq2SeqLM.from_pretrained(
|
13 |
+
model_name, use_auth_token=access_token
|
14 |
+
)
|
15 |
+
tokenizer = AutoTokenizer.from_pretrained(
|
16 |
+
model_name, use_auth_token=access_token
|
17 |
+
)
|
18 |
+
|
19 |
+
bert_tokenizer = AutoTokenizer.from_pretrained(
|
20 |
+
bert_model_name, use_auth_token=access_token
|
21 |
+
)
|
22 |
+
bert_model = AutoModelForSequenceClassification.from_pretrained(
|
23 |
+
bert_model_name, use_auth_token=access_token
|
24 |
+
)
|
25 |
+
|
26 |
+
|
27 |
+
def correlation_score(table, review):
|
28 |
+
# Compute the correlation score
|
29 |
+
args = ((table, review))
|
30 |
+
inputs = bert_tokenizer(*args, padding=True, max_length=128, truncation=True, return_tensors="pt")
|
31 |
+
logits = model(**inputs).logits
|
32 |
+
probs = logits.softmax(dim=-1)
|
33 |
+
return {
|
34 |
+
"correlation": probs[:, 1].item()
|
35 |
+
}
|
36 |
|
37 |
def create_prompt(stars, useful, funny, cool):
|
38 |
return f"Generate review: stars: {stars}, useful: {useful}, funny: {funny}, cool: {cool}"
|
39 |
|
40 |
+
|
41 |
def postprocess(review):
|
42 |
dot = review.rfind('.')
|
43 |
return review[:dot+1]
|
44 |
|
45 |
+
|
46 |
def generate_reviews(stars, useful, funny, cool):
|
47 |
text = create_prompt(stars, useful, funny, cool)
|
48 |
inputs = tokenizer(text, return_tensors='pt')
|
|
|
55 |
top_p=0.9
|
56 |
)
|
57 |
reviews = []
|
58 |
+
scores = []
|
59 |
for review in out:
|
60 |
reviews.append(postprocess(tokenizer.decode(review, skip_special_tokens=True)))
|
61 |
+
scores.append(
|
62 |
+
correlation_score(text[17:], review)
|
63 |
+
)
|
64 |
+
|
65 |
+
return reviews[0], reviews[1], reviews[2], scores[0], scores[1], scores[2]
|
66 |
|
|
|
67 |
|
68 |
css = """
|
69 |
#ctr {text-align: center;}
|
|
|
95 |
with demo:
|
96 |
with gr.Row():
|
97 |
gr.Markdown(md_text)
|
98 |
+
|
99 |
with gr.Row():
|
100 |
+
stars = gr.inputs.Slider(minimum=0, maximum=5,
|
101 |
+
step=1, default=0, label="stars")
|
102 |
+
useful = gr.inputs.Slider(
|
103 |
+
minimum=0, maximum=5, step=1, default=0, label="useful")
|
104 |
+
funny = gr.inputs.Slider(minimum=0, maximum=5,
|
105 |
+
step=1, default=0, label="funny")
|
106 |
+
cool = gr.inputs.Slider(minimum=0, maximum=5,
|
107 |
+
step=1, default=0, label="cool")
|
108 |
with gr.Row():
|
109 |
button = gr.Button("Generate reviews !", elem_id='btn')
|
110 |
|
|
|
113 |
output2 = gr.Textbox(label="Review #2")
|
114 |
output3 = gr.Textbox(label="Review #3")
|
115 |
|
116 |
+
with gr.Row():
|
117 |
+
score1 = gr.Label(label="Correlation score #1")
|
118 |
+
score2 = gr.Label(label="Correlation score #2")
|
119 |
+
score3 = gr.Label(label="Correlation score #3")
|
120 |
+
|
121 |
with gr.Row():
|
122 |
gr.Markdown(resources)
|
123 |
|
124 |
button.click(
|
125 |
fn=generate_reviews,
|
126 |
inputs=[stars, useful, funny, cool],
|
127 |
+
outputs=[output1, output2, output3, score1, score2, score3]
|
128 |
)
|
129 |
|
130 |
+
demo.launch()
|