nanhsin commited on
Commit
864a52e
·
verified ·
1 Parent(s): 1af09ab

Paste Question 4 for Lab 7

Browse files

Deploy an interactive chart for the moving average of the approval rate for Joe Biden

Files changed (1) hide show
  1. app.py +56 -137
app.py CHANGED
@@ -1,147 +1,66 @@
1
- import io
2
- import random
3
- from typing import List, Tuple
4
-
5
- import aiohttp
6
  import panel as pn
7
- from PIL import Image
8
- from transformers import CLIPModel, CLIPProcessor
9
-
10
- pn.extension(design="bootstrap", sizing_mode="stretch_width")
11
-
12
- ICON_URLS = {
13
- "brand-github": "https://github.com/holoviz/panel",
14
- "brand-twitter": "https://twitter.com/Panel_Org",
15
- "brand-linkedin": "https://www.linkedin.com/company/panel-org",
16
- "message-circle": "https://discourse.holoviz.org/",
17
- "brand-discord": "https://discord.gg/AXRHnJU6sP",
18
- }
19
-
20
 
21
- async def random_url(_):
22
- pet = random.choice(["cat", "dog"])
23
- api_url = f"https://api.the{pet}api.com/v1/images/search"
24
- async with aiohttp.ClientSession() as session:
25
- async with session.get(api_url) as resp:
26
- return (await resp.json())[0]["url"]
27
-
28
-
29
- @pn.cache
30
- def load_processor_model(
31
- processor_name: str, model_name: str
32
- ) -> Tuple[CLIPProcessor, CLIPModel]:
33
- processor = CLIPProcessor.from_pretrained(processor_name)
34
- model = CLIPModel.from_pretrained(model_name)
35
- return processor, model
36
 
 
37
 
38
- async def open_image_url(image_url: str) -> Image:
39
- async with aiohttp.ClientSession() as session:
40
- async with session.get(image_url) as resp:
41
- return Image.open(io.BytesIO(await resp.read()))
42
 
 
 
 
 
 
 
43
 
44
- def get_similarity_scores(class_items: List[str], image: Image) -> List[float]:
45
- processor, model = load_processor_model(
46
- "openai/clip-vit-base-patch32", "openai/clip-vit-base-patch32"
 
47
  )
48
- inputs = processor(
49
- text=class_items,
50
- images=[image],
51
- return_tensors="pt", # pytorch tensors
52
- )
53
- outputs = model(**inputs)
54
- logits_per_image = outputs.logits_per_image
55
- class_likelihoods = logits_per_image.softmax(dim=1).detach().numpy()
56
- return class_likelihoods[0]
57
 
 
 
 
 
 
58
 
59
- async def process_inputs(class_names: List[str], image_url: str):
60
- """
61
- High level function that takes in the user inputs and returns the
62
- classification results as panel objects.
63
- """
64
- try:
65
- main.disabled = True
66
- if not image_url:
67
- yield "##### ⚠️ Provide an image URL"
68
- return
69
-
70
- yield "##### ⚙ Fetching image and running model..."
71
- try:
72
- pil_img = await open_image_url(image_url)
73
- img = pn.pane.Image(pil_img, height=400, align="center")
74
- except Exception as e:
75
- yield f"##### 😔 Something went wrong, please try a different URL!"
76
- return
77
 
78
- class_items = class_names.split(",")
79
- class_likelihoods = get_similarity_scores(class_items, pil_img)
80
-
81
- # build the results column
82
- results = pn.Column("##### 🎉 Here are the results!", img)
83
-
84
- for class_item, class_likelihood in zip(class_items, class_likelihoods):
85
- row_label = pn.widgets.StaticText(
86
- name=class_item.strip(), value=f"{class_likelihood:.2%}", align="center"
87
- )
88
- row_bar = pn.indicators.Progress(
89
- value=int(class_likelihood * 100),
90
- sizing_mode="stretch_width",
91
- bar_color="secondary",
92
- margin=(0, 10),
93
- design=pn.theme.Material,
94
- )
95
- results.append(pn.Column(row_label, row_bar))
96
- yield results
97
- finally:
98
- main.disabled = False
99
-
100
-
101
- # create widgets
102
- randomize_url = pn.widgets.Button(name="Randomize URL", align="end")
103
-
104
- image_url = pn.widgets.TextInput(
105
- name="Image URL to classify",
106
- value=pn.bind(random_url, randomize_url),
107
- )
108
- class_names = pn.widgets.TextInput(
109
- name="Comma separated class names",
110
- placeholder="Enter possible class names, e.g. cat, dog",
111
- value="cat, dog, parrot",
112
- )
113
-
114
- input_widgets = pn.Column(
115
- "##### 😊 Click randomize or paste a URL to start classifying!",
116
- pn.Row(image_url, randomize_url),
117
- class_names,
118
- )
119
-
120
- # add interactivity
121
- interactive_result = pn.panel(
122
- pn.bind(process_inputs, image_url=image_url, class_names=class_names),
123
- height=600,
124
- )
125
-
126
- # add footer
127
- footer_row = pn.Row(pn.Spacer(), align="center")
128
- for icon, url in ICON_URLS.items():
129
- href_button = pn.widgets.Button(icon=icon, width=35, height=35)
130
- href_button.js_on_click(code=f"window.open('{url}')")
131
- footer_row.append(href_button)
132
- footer_row.append(pn.Spacer())
133
-
134
- # create dashboard
135
- main = pn.WidgetBox(
136
- input_widgets,
137
- interactive_result,
138
- footer_row,
139
- )
140
-
141
- title = "Panel Demo - Image Classification"
142
- pn.template.BootstrapTemplate(
143
- title=title,
144
- main=main,
145
- main_max_width="min(50%, 698px)",
146
- header_background="#F08080",
147
- ).servable(title=title)
 
 
 
 
 
 
1
  import panel as pn
2
+ import vega_datasets
 
 
 
 
 
 
 
 
 
 
 
 
3
 
4
+ # Enable Panel extensions
5
+ pn.extension(design='bootstrap')
6
+ pn.extension('vega')
7
+ template = pn.template.BootstrapTemplate(
8
+ title='Nan-Hsin Lin',
9
+ )
 
 
 
 
 
 
 
 
 
10
 
11
+ # Define a function to create and return a plot
12
 
13
+ def create_plot(subgroup, date_range, moving_av_window):
 
 
 
14
 
15
+ # Apply any required transformations to the data in pandas
16
+ df3 = df2[df2['choice'] == 'approve'].copy()
17
+ df3 = df3[df3['subgroup'] == subgroup]
18
+ df3['smoothed_rate'] = df3['rate'].rolling(moving_av_window).mean().shift(-int(moving_av_window/2))
19
+ start_date, end_date = date_range
20
+ df3 = df3[(df3['timestamp'] >= np.datetime64(start_date)) & (df3['timestamp'] <= np.datetime64(end_date))]
21
 
22
+ # Line chart
23
+ rate_line = alt.Chart(df3).mark_line(strokeWidth=2, color='red').encode(
24
+ x=alt.X('timestamp:T', axis=alt.Axis(title=None)),
25
+ y=alt.Y('average(smoothed_rate):Q', axis=alt.Axis(title='move_avg'), scale=alt.Scale(domain=[30, 60]))
26
  )
 
 
 
 
 
 
 
 
 
27
 
28
+ # Scatter plot with individual polls
29
+ rate_scatter = alt.Chart(df3).mark_point(color='grey', size=2, opacity=0.7).encode(
30
+ x=alt.X('timestamp:T', axis=alt.Axis(title=None)),
31
+ y=alt.Y('average(rate):Q', axis=alt.Axis(title='approve'), scale=alt.Scale(domain=[30, 60])),
32
+ )
33
 
34
+ # Put them together
35
+ plot = alt.layer(rate_line, rate_scatter).configure_view(strokeWidth=0)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
36
 
37
+ # Return the combined chart
38
+ return plot
39
+
40
+ # Create the selection widget
41
+ select = pn.widgets.Select(name='Select', options=df2['subgroup'].unique().tolist())
42
+
43
+ # Create the slider for the date range
44
+ dateSlider = pn.widgets.DateRangeSlider(name='Date Range Slider',
45
+ start=df2['timestamp'].min(),
46
+ end=df2['timestamp'].max(),
47
+ value=(df2['timestamp'].min(), df2['timestamp'].max()))
48
+
49
+ # Create the slider for the moving average window
50
+ avgSlider = pn.widgets.IntSlider(name='Moving average window', start=1, end=100, value=1)
51
+
52
+ # Bind the widgets to the create_plot function
53
+ plot_widgets = pn.Row(pn.bind(create_plot, select, dateSlider, avgSlider))
54
+
55
+ # Combine everything in a Panel Column to create an app
56
+ maincol = pn.Column()
57
+ maincol.append("# SI649 Lab07")
58
+ maincol.append("Hello! This is **Nan**. I love information visualization! Email me: [[email protected]](mailto:[email protected])")
59
+ maincol.append(plot_widgets)
60
+ maincol.append(select)
61
+ maincol.append(dateSlider)
62
+ maincol.append(avgSlider)
63
+ template.main.append(maincol)
64
+
65
+ # set the app to be servable
66
+ template.servable(title="Nan-Hsin Lin")