Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,169 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# https://scikit-learn.org/stable/auto_examples/cluster/plot_birch_vs_minibatchkmeans.html
|
2 |
+
|
3 |
+
from itertools import cycle
|
4 |
+
from time import time
|
5 |
+
|
6 |
+
import gradio as gr
|
7 |
+
import matplotlib.colors as colors
|
8 |
+
import matplotlib.pyplot as plt
|
9 |
+
import numpy as np
|
10 |
+
from joblib import cpu_count
|
11 |
+
from sklearn.cluster import Birch, MiniBatchKMeans
|
12 |
+
from sklearn.datasets import make_blobs
|
13 |
+
|
14 |
+
plt.switch_backend("agg")
|
15 |
+
|
16 |
+
|
17 |
+
def do_submit(n_samples, birch_threshold, birch_n_clusters):
|
18 |
+
n_samples = int(n_samples)
|
19 |
+
birch_threshold = float(birch_threshold)
|
20 |
+
birch_n_clusters = int(birch_n_clusters)
|
21 |
+
result = ""
|
22 |
+
|
23 |
+
# Generate centers for the blobs so that it forms a 10 X 10 grid.
|
24 |
+
xx = np.linspace(-22, 22, 10)
|
25 |
+
yy = np.linspace(-22, 22, 10)
|
26 |
+
xx, yy = np.meshgrid(xx, yy)
|
27 |
+
n_centers = np.hstack((np.ravel(xx)[:, np.newaxis], np.ravel(yy)[:, np.newaxis]))
|
28 |
+
|
29 |
+
# Generate blobs to do a comparison between MiniBatchKMeans and BIRCH.
|
30 |
+
X, y = make_blobs(n_samples=n_samples, centers=n_centers, random_state=0)
|
31 |
+
|
32 |
+
# Use all colors that matplotlib provides by default.
|
33 |
+
colors_ = cycle(colors.cnames.keys())
|
34 |
+
|
35 |
+
fig = plt.figure(figsize=(12, 4))
|
36 |
+
fig.subplots_adjust(left=0.04, right=0.98, bottom=0.1, top=0.9)
|
37 |
+
|
38 |
+
# Compute clustering with BIRCH with and without the final clustering step
|
39 |
+
# and plot.
|
40 |
+
birch_models = [
|
41 |
+
Birch(threshold=1.7, n_clusters=None),
|
42 |
+
Birch(threshold=1.7, n_clusters=100),
|
43 |
+
]
|
44 |
+
final_step = ["without global clustering", "with global clustering"]
|
45 |
+
|
46 |
+
for ind, (birch_model, info) in enumerate(zip(birch_models, final_step)):
|
47 |
+
t = time()
|
48 |
+
birch_model.fit(X)
|
49 |
+
result += (
|
50 |
+
"BIRCH %s as the final step took %0.2f seconds" % (info, (time() - t))
|
51 |
+
+ "\n"
|
52 |
+
)
|
53 |
+
|
54 |
+
# Plot result
|
55 |
+
labels = birch_model.labels_
|
56 |
+
centroids = birch_model.subcluster_centers_
|
57 |
+
n_clusters = np.unique(labels).size
|
58 |
+
result = result + "n_clusters : %d" % n_clusters + "\n"
|
59 |
+
|
60 |
+
ax = fig.add_subplot(1, 3, ind + 1)
|
61 |
+
for this_centroid, k, col in zip(centroids, range(n_clusters), colors_):
|
62 |
+
mask = labels == k
|
63 |
+
ax.scatter(
|
64 |
+
X[mask, 0], X[mask, 1], c="w", edgecolor=col, marker=".", alpha=0.5
|
65 |
+
)
|
66 |
+
if birch_model.n_clusters is None:
|
67 |
+
ax.scatter(this_centroid[0], this_centroid[1], marker="+", c="k", s=25)
|
68 |
+
ax.set_ylim([-25, 25])
|
69 |
+
ax.set_xlim([-25, 25])
|
70 |
+
ax.set_autoscaley_on(False)
|
71 |
+
ax.set_title("BIRCH %s" % info)
|
72 |
+
|
73 |
+
# Compute clustering with MiniBatchKMeans.
|
74 |
+
mbk = MiniBatchKMeans(
|
75 |
+
init="k-means++",
|
76 |
+
n_clusters=100,
|
77 |
+
batch_size=256 * cpu_count(),
|
78 |
+
n_init=10,
|
79 |
+
max_no_improvement=10,
|
80 |
+
verbose=0,
|
81 |
+
random_state=0,
|
82 |
+
)
|
83 |
+
t0 = time()
|
84 |
+
mbk.fit(X)
|
85 |
+
t_mini_batch = time() - t0
|
86 |
+
result += "Time taken to run MiniBatchKMeans %0.2f seconds" % t_mini_batch + "\n"
|
87 |
+
mbk_means_labels_unique = np.unique(mbk.labels_)
|
88 |
+
|
89 |
+
ax = fig.add_subplot(1, 3, 3)
|
90 |
+
for this_centroid, k, col in zip(mbk.cluster_centers_, range(n_clusters), colors_):
|
91 |
+
mask = mbk.labels_ == k
|
92 |
+
ax.scatter(X[mask, 0], X[mask, 1], marker=".", c="w", edgecolor=col, alpha=0.5)
|
93 |
+
ax.scatter(this_centroid[0], this_centroid[1], marker="+", c="k", s=25)
|
94 |
+
ax.set_xlim([-25, 25])
|
95 |
+
ax.set_ylim([-25, 25])
|
96 |
+
ax.set_title("MiniBatchKMeans")
|
97 |
+
ax.set_autoscaley_on(False)
|
98 |
+
|
99 |
+
return fig, result
|
100 |
+
|
101 |
+
|
102 |
+
# Theme from - https://huggingface.co/spaces/trl-lib/stack-llama/blob/main/app.py
|
103 |
+
theme = gr.themes.Monochrome(
|
104 |
+
primary_hue="indigo",
|
105 |
+
secondary_hue="blue",
|
106 |
+
neutral_hue="slate",
|
107 |
+
radius_size=gr.themes.sizes.radius_sm,
|
108 |
+
font=[
|
109 |
+
gr.themes.GoogleFont("Open Sans"),
|
110 |
+
"ui-sans-serif",
|
111 |
+
"system-ui",
|
112 |
+
"sans-serif",
|
113 |
+
],
|
114 |
+
)
|
115 |
+
|
116 |
+
title = "Compare BIRCH and MiniBatchKMeans"
|
117 |
+
with gr.Blocks(title=title, theme=theme) as demo:
|
118 |
+
gr.Markdown(f"## {title}")
|
119 |
+
gr.Markdown(
|
120 |
+
"[Scikit-learn Example](https://scikit-learn.org/stable/auto_examples/cluster/plot_birch_vs_minibatchkmeans.html)"
|
121 |
+
)
|
122 |
+
|
123 |
+
gr.Markdown(
|
124 |
+
"This example compares the timing of BIRCH (with and without the global clustering step) and \
|
125 |
+
MiniBatchKMeans on a synthetic dataset having 25,000 samples and 2 features generated using make_blobs.\
|
126 |
+
\n Both MiniBatchKMeans and BIRCH are very scalable algorithms and could run efficiently on hundreds of thousands or \
|
127 |
+
even millions of datapoints. We chose to limit the dataset size of this example in the interest of keeping our \
|
128 |
+
Continuous Integration resource usage reasonable but the interested reader might enjoy editing this script to \
|
129 |
+
rerun it with a larger value for n_samples.\
|
130 |
+
\n\n\
|
131 |
+
If n_clusters is set to None, the data is reduced from 25,000 samples to a set of 158 clusters. This can be viewed as a preprocessing step before the final (global) clustering step that further reduces these 158 clusters to 100 clusters."
|
132 |
+
)
|
133 |
+
|
134 |
+
n_samples = gr.Slider(
|
135 |
+
minimum=20000,
|
136 |
+
maximum=80000,
|
137 |
+
label="Number of samples",
|
138 |
+
step=500,
|
139 |
+
value=25000,
|
140 |
+
)
|
141 |
+
birch_threshold = gr.Slider(
|
142 |
+
minimum=0.5,
|
143 |
+
maximum=2.0,
|
144 |
+
label="Birch Threshold",
|
145 |
+
step=0.1,
|
146 |
+
value=1.7,
|
147 |
+
)
|
148 |
+
birch_n_clusters = gr.Slider(
|
149 |
+
minimum=0,
|
150 |
+
maximum=100,
|
151 |
+
label="Birch number of clusters",
|
152 |
+
step=1,
|
153 |
+
value=100,
|
154 |
+
)
|
155 |
+
|
156 |
+
plt_out = gr.Plot()
|
157 |
+
output = gr.Textbox(label="Output", multiline=True)
|
158 |
+
|
159 |
+
sub_btn = gr.Button("Submit")
|
160 |
+
sub_btn.click(
|
161 |
+
fn=do_submit,
|
162 |
+
inputs=[n_samples, birch_threshold, birch_n_clusters],
|
163 |
+
outputs=[plt_out, output],
|
164 |
+
)
|
165 |
+
|
166 |
+
|
167 |
+
if __name__ == "__main__":
|
168 |
+
demo.launch()
|
169 |
+
|