Spaces:
Running
Running
MilesCranmer
commited on
Commit
•
a206d6a
1
Parent(s):
ef7aada
refactor(gui): gradio to use object oriented wrapper
Browse files- gui/app.py +162 -136
- gui/data.py +1 -1
- gui/processing.py +2 -0
gui/app.py
CHANGED
@@ -1,56 +1,78 @@
|
|
|
|
|
|
1 |
import gradio as gr
|
2 |
-
|
|
|
|
|
3 |
from plots import plot_example_data, plot_pareto_curve
|
4 |
from processing import processing
|
5 |
|
6 |
-
GLOBAL_SETTINGS = dict(theme="default")
|
7 |
-
|
8 |
|
9 |
-
|
10 |
-
|
11 |
-
# Plot of the example data:
|
12 |
with gr.Row():
|
|
|
13 |
with gr.Column():
|
14 |
-
example_plot = gr.Plot()
|
15 |
with gr.Column():
|
16 |
-
test_equation = gr.Radio(
|
17 |
-
|
18 |
)
|
19 |
-
num_points = gr.Slider(
|
20 |
minimum=10,
|
21 |
maximum=1000,
|
22 |
value=200,
|
23 |
label="Number of Data Points",
|
24 |
step=1,
|
25 |
)
|
26 |
-
noise_level = gr.Slider(
|
27 |
minimum=0, maximum=1, value=0.05, label="Noise Level"
|
28 |
)
|
29 |
-
data_seed = gr.Number(value=0, label="Random Seed")
|
30 |
-
|
31 |
-
|
32 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
"The rightmost column of your CSV file will be used as the target variable."
|
34 |
)
|
35 |
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
)
|
44 |
|
45 |
|
46 |
-
|
47 |
-
|
48 |
-
binary_operators = gr.CheckboxGroup(
|
49 |
choices=["+", "-", "*", "/", "^", "max", "min", "mod", "cond"],
|
50 |
label="Binary Operators",
|
51 |
value=["+", "-", "*", "/"],
|
52 |
)
|
53 |
-
unary_operators = gr.CheckboxGroup(
|
54 |
choices=[
|
55 |
"sin",
|
56 |
"cos",
|
@@ -69,58 +91,61 @@ def _settings_layout():
|
|
69 |
label="Unary Operators",
|
70 |
value=["sin"],
|
71 |
)
|
72 |
-
niterations = gr.Slider(
|
73 |
minimum=1,
|
74 |
maximum=1000,
|
75 |
value=40,
|
76 |
label="Number of Iterations",
|
77 |
step=1,
|
78 |
)
|
79 |
-
maxsize = gr.Slider(
|
80 |
minimum=7,
|
81 |
maximum=100,
|
82 |
value=20,
|
83 |
label="Maximum Complexity",
|
84 |
step=1,
|
85 |
)
|
86 |
-
parsimony = gr.Number(
|
87 |
value=0.0032,
|
88 |
label="Parsimony Coefficient",
|
89 |
)
|
90 |
-
|
91 |
-
|
|
|
|
|
|
|
92 |
minimum=2,
|
93 |
maximum=100,
|
94 |
value=15,
|
95 |
label="Number of Populations",
|
96 |
step=1,
|
97 |
)
|
98 |
-
population_size = gr.Slider(
|
99 |
minimum=2,
|
100 |
maximum=1000,
|
101 |
value=33,
|
102 |
label="Population Size",
|
103 |
step=1,
|
104 |
)
|
105 |
-
ncycles_per_iteration = gr.Number(
|
106 |
value=550,
|
107 |
label="Cycles per Iteration",
|
108 |
)
|
109 |
-
elementwise_loss = gr.Radio(
|
110 |
["L2DistLoss()", "L1DistLoss()", "LogitDistLoss()", "HuberLoss()"],
|
111 |
value="L2DistLoss()",
|
112 |
label="Loss Function",
|
113 |
)
|
114 |
-
adaptive_parsimony_scaling = gr.Number(
|
115 |
value=20.0,
|
116 |
label="Adaptive Parsimony Scaling",
|
117 |
)
|
118 |
-
optimizer_algorithm = gr.Radio(
|
119 |
["BFGS", "NelderMead"],
|
120 |
value="BFGS",
|
121 |
label="Optimizer Algorithm",
|
122 |
)
|
123 |
-
optimizer_iterations = gr.Slider(
|
124 |
minimum=1,
|
125 |
maximum=100,
|
126 |
value=8,
|
@@ -128,11 +153,11 @@ def _settings_layout():
|
|
128 |
step=1,
|
129 |
)
|
130 |
# Bool:
|
131 |
-
batching = gr.Checkbox(
|
132 |
value=False,
|
133 |
label="Batching",
|
134 |
)
|
135 |
-
batch_size = gr.Slider(
|
136 |
minimum=2,
|
137 |
maximum=1000,
|
138 |
value=50,
|
@@ -140,121 +165,122 @@ def _settings_layout():
|
|
140 |
step=1,
|
141 |
)
|
142 |
|
143 |
-
|
144 |
-
|
|
|
|
|
145 |
minimum=1,
|
146 |
maximum=100,
|
147 |
value=3,
|
148 |
label="Plot Update Delay",
|
149 |
)
|
150 |
-
force_run = gr.Checkbox(
|
151 |
value=False,
|
152 |
label="Ignore Warnings",
|
153 |
)
|
154 |
-
|
155 |
-
|
156 |
-
|
157 |
-
|
158 |
-
|
159 |
-
|
160 |
-
|
161 |
-
|
162 |
-
|
163 |
-
|
164 |
-
|
165 |
-
|
166 |
-
|
167 |
-
|
168 |
-
|
169 |
-
|
170 |
-
|
171 |
-
|
172 |
-
|
173 |
-
|
174 |
-
|
175 |
-
|
176 |
-
|
177 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
178 |
with gr.Row():
|
179 |
with gr.Column():
|
180 |
with gr.Row():
|
181 |
-
|
182 |
with gr.Row():
|
183 |
-
|
184 |
-
|
185 |
with gr.Column():
|
186 |
-
|
187 |
-
|
188 |
-
with gr.Tab("Predictions"):
|
189 |
-
blocks["predictions_plot"] = gr.Plot()
|
190 |
-
|
191 |
-
blocks["df"] = gr.Dataframe(
|
192 |
-
headers=["complexity", "loss", "equation"],
|
193 |
-
datatype=["number", "number", "str"],
|
194 |
-
wrap=True,
|
195 |
-
column_widths=[75, 75, 200],
|
196 |
-
interactive=False,
|
197 |
-
)
|
198 |
-
blocks["run"] = gr.Button()
|
199 |
-
|
200 |
-
blocks["run"].click(
|
201 |
-
processing,
|
202 |
-
inputs=[
|
203 |
-
blocks[k]
|
204 |
-
for k in [
|
205 |
-
"file_input",
|
206 |
-
"force_run",
|
207 |
-
"test_equation",
|
208 |
-
"num_points",
|
209 |
-
"noise_level",
|
210 |
-
"data_seed",
|
211 |
-
"niterations",
|
212 |
-
"maxsize",
|
213 |
-
"binary_operators",
|
214 |
-
"unary_operators",
|
215 |
-
"plot_update_delay",
|
216 |
-
"parsimony",
|
217 |
-
"populations",
|
218 |
-
"population_size",
|
219 |
-
"ncycles_per_iteration",
|
220 |
-
"elementwise_loss",
|
221 |
-
"adaptive_parsimony_scaling",
|
222 |
-
"optimizer_algorithm",
|
223 |
-
"optimizer_iterations",
|
224 |
-
"batching",
|
225 |
-
"batch_size",
|
226 |
-
]
|
227 |
-
],
|
228 |
-
outputs=[blocks["df"], blocks["predictions_plot"]],
|
229 |
-
show_progress=True,
|
230 |
-
)
|
231 |
-
|
232 |
-
# Any update to the equation choice will trigger a plot_example_data:
|
233 |
-
eqn_components = [
|
234 |
-
blocks["test_equation"],
|
235 |
-
blocks["num_points"],
|
236 |
-
blocks["noise_level"],
|
237 |
-
blocks["data_seed"],
|
238 |
-
]
|
239 |
-
for eqn_component in eqn_components:
|
240 |
-
eqn_component.change(
|
241 |
-
plot_example_data,
|
242 |
-
eqn_components,
|
243 |
-
blocks["example_plot"],
|
244 |
-
show_progress=False,
|
245 |
-
)
|
246 |
|
247 |
# Update plot when dataframe is updated:
|
248 |
-
|
249 |
plot_pareto_curve,
|
250 |
-
inputs=[
|
251 |
-
outputs=[
|
252 |
show_progress=False,
|
253 |
)
|
254 |
-
demo.load(plot_example_data, eqn_components, blocks["example_plot"])
|
255 |
|
256 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
257 |
|
258 |
|
259 |
if __name__ == "__main__":
|
260 |
-
|
|
|
1 |
+
from collections import OrderedDict
|
2 |
+
|
3 |
import gradio as gr
|
4 |
+
import numpy as np
|
5 |
+
from data import TEST_EQUATIONS
|
6 |
+
from gradio.components.base import Component
|
7 |
from plots import plot_example_data, plot_pareto_curve
|
8 |
from processing import processing
|
9 |
|
|
|
|
|
10 |
|
11 |
+
class ExampleData:
|
12 |
+
def __init__(self, demo: gr.Blocks) -> None:
|
|
|
13 |
with gr.Row():
|
14 |
+
# Plot of the example data:
|
15 |
with gr.Column():
|
16 |
+
self.example_plot = gr.Plot()
|
17 |
with gr.Column():
|
18 |
+
self.test_equation = gr.Radio(
|
19 |
+
TEST_EQUATIONS, value=TEST_EQUATIONS[0], label="Test Equation"
|
20 |
)
|
21 |
+
self.num_points = gr.Slider(
|
22 |
minimum=10,
|
23 |
maximum=1000,
|
24 |
value=200,
|
25 |
label="Number of Data Points",
|
26 |
step=1,
|
27 |
)
|
28 |
+
self.noise_level = gr.Slider(
|
29 |
minimum=0, maximum=1, value=0.05, label="Noise Level"
|
30 |
)
|
31 |
+
self.data_seed = gr.Number(value=0, label="Random Seed")
|
32 |
+
|
33 |
+
# Set up plotting:
|
34 |
+
|
35 |
+
eqn_components = [
|
36 |
+
self.test_equation,
|
37 |
+
self.num_points,
|
38 |
+
self.noise_level,
|
39 |
+
self.data_seed,
|
40 |
+
]
|
41 |
+
for eqn_component in eqn_components:
|
42 |
+
eqn_component.change(
|
43 |
+
plot_example_data,
|
44 |
+
eqn_components,
|
45 |
+
self.example_plot,
|
46 |
+
show_progress=False,
|
47 |
+
)
|
48 |
+
|
49 |
+
demo.load(plot_example_data, eqn_components, self.example_plot)
|
50 |
+
|
51 |
+
|
52 |
+
class UploadData:
|
53 |
+
def __init__(self) -> None:
|
54 |
+
self.file_input = gr.File(label="Upload a CSV File")
|
55 |
+
self.label = gr.Markdown(
|
56 |
"The rightmost column of your CSV file will be used as the target variable."
|
57 |
)
|
58 |
|
59 |
+
|
60 |
+
class Data:
|
61 |
+
def __init__(self, demo: gr.Blocks) -> None:
|
62 |
+
with gr.Tab("Example Data"):
|
63 |
+
self.example_data = ExampleData(demo)
|
64 |
+
with gr.Tab("Upload Data"):
|
65 |
+
self.upload_data = UploadData()
|
|
|
66 |
|
67 |
|
68 |
+
class BasicSettings:
|
69 |
+
def __init__(self) -> None:
|
70 |
+
self.binary_operators = gr.CheckboxGroup(
|
71 |
choices=["+", "-", "*", "/", "^", "max", "min", "mod", "cond"],
|
72 |
label="Binary Operators",
|
73 |
value=["+", "-", "*", "/"],
|
74 |
)
|
75 |
+
self.unary_operators = gr.CheckboxGroup(
|
76 |
choices=[
|
77 |
"sin",
|
78 |
"cos",
|
|
|
91 |
label="Unary Operators",
|
92 |
value=["sin"],
|
93 |
)
|
94 |
+
self.niterations = gr.Slider(
|
95 |
minimum=1,
|
96 |
maximum=1000,
|
97 |
value=40,
|
98 |
label="Number of Iterations",
|
99 |
step=1,
|
100 |
)
|
101 |
+
self.maxsize = gr.Slider(
|
102 |
minimum=7,
|
103 |
maximum=100,
|
104 |
value=20,
|
105 |
label="Maximum Complexity",
|
106 |
step=1,
|
107 |
)
|
108 |
+
self.parsimony = gr.Number(
|
109 |
value=0.0032,
|
110 |
label="Parsimony Coefficient",
|
111 |
)
|
112 |
+
|
113 |
+
|
114 |
+
class AdvancedSettings:
|
115 |
+
def __init__(self) -> None:
|
116 |
+
self.populations = gr.Slider(
|
117 |
minimum=2,
|
118 |
maximum=100,
|
119 |
value=15,
|
120 |
label="Number of Populations",
|
121 |
step=1,
|
122 |
)
|
123 |
+
self.population_size = gr.Slider(
|
124 |
minimum=2,
|
125 |
maximum=1000,
|
126 |
value=33,
|
127 |
label="Population Size",
|
128 |
step=1,
|
129 |
)
|
130 |
+
self.ncycles_per_iteration = gr.Number(
|
131 |
value=550,
|
132 |
label="Cycles per Iteration",
|
133 |
)
|
134 |
+
self.elementwise_loss = gr.Radio(
|
135 |
["L2DistLoss()", "L1DistLoss()", "LogitDistLoss()", "HuberLoss()"],
|
136 |
value="L2DistLoss()",
|
137 |
label="Loss Function",
|
138 |
)
|
139 |
+
self.adaptive_parsimony_scaling = gr.Number(
|
140 |
value=20.0,
|
141 |
label="Adaptive Parsimony Scaling",
|
142 |
)
|
143 |
+
self.optimizer_algorithm = gr.Radio(
|
144 |
["BFGS", "NelderMead"],
|
145 |
value="BFGS",
|
146 |
label="Optimizer Algorithm",
|
147 |
)
|
148 |
+
self.optimizer_iterations = gr.Slider(
|
149 |
minimum=1,
|
150 |
maximum=100,
|
151 |
value=8,
|
|
|
153 |
step=1,
|
154 |
)
|
155 |
# Bool:
|
156 |
+
self.batching = gr.Checkbox(
|
157 |
value=False,
|
158 |
label="Batching",
|
159 |
)
|
160 |
+
self.batch_size = gr.Slider(
|
161 |
minimum=2,
|
162 |
maximum=1000,
|
163 |
value=50,
|
|
|
165 |
step=1,
|
166 |
)
|
167 |
|
168 |
+
|
169 |
+
class GradioSettings:
|
170 |
+
def __init__(self) -> None:
|
171 |
+
self.plot_update_delay = gr.Slider(
|
172 |
minimum=1,
|
173 |
maximum=100,
|
174 |
value=3,
|
175 |
label="Plot Update Delay",
|
176 |
)
|
177 |
+
self.force_run = gr.Checkbox(
|
178 |
value=False,
|
179 |
label="Ignore Warnings",
|
180 |
)
|
181 |
+
|
182 |
+
|
183 |
+
class Settings:
|
184 |
+
def __init__(self):
|
185 |
+
with gr.Tab("Basic Settings"):
|
186 |
+
self.basic_settings = BasicSettings()
|
187 |
+
with gr.Tab("Advanced Settings"):
|
188 |
+
self.advanced_settings = AdvancedSettings()
|
189 |
+
with gr.Tab("Gradio Settings"):
|
190 |
+
self.gradio_settings = GradioSettings()
|
191 |
+
|
192 |
+
|
193 |
+
class Results:
|
194 |
+
def __init__(self):
|
195 |
+
with gr.Tab("Pareto Front"):
|
196 |
+
self.pareto = gr.Plot()
|
197 |
+
with gr.Tab("Predictions"):
|
198 |
+
self.predictions_plot = gr.Plot()
|
199 |
+
|
200 |
+
self.df = gr.Dataframe(
|
201 |
+
headers=["complexity", "loss", "equation"],
|
202 |
+
datatype=["number", "number", "str"],
|
203 |
+
wrap=True,
|
204 |
+
column_widths=[75, 75, 200],
|
205 |
+
interactive=False,
|
206 |
+
)
|
207 |
+
|
208 |
+
|
209 |
+
def flatten_attributes(component_group, absolute_name: str, d=None) -> OrderedDict:
|
210 |
+
if d is None:
|
211 |
+
d = OrderedDict()
|
212 |
+
|
213 |
+
if not hasattr(component_group, "__dict__"):
|
214 |
+
return d
|
215 |
+
|
216 |
+
for name, elem in component_group.__dict__.items():
|
217 |
+
new_absolute_name = absolute_name + "." + name
|
218 |
+
if name.startswith("_"):
|
219 |
+
# Private attribute
|
220 |
+
continue
|
221 |
+
elif elem in component_group.__dict__.values():
|
222 |
+
# Don't duplicate any tiems
|
223 |
+
continue
|
224 |
+
elif isinstance(elem, Component):
|
225 |
+
# Only add components to dict
|
226 |
+
d[new_absolute_name] = elem
|
227 |
+
else:
|
228 |
+
d = flatten_attributes(elem, new_absolute_name, d=d)
|
229 |
+
|
230 |
+
return d
|
231 |
+
|
232 |
+
|
233 |
+
class AppInterface:
|
234 |
+
def __init__(self, demo: gr.Blocks) -> None:
|
235 |
with gr.Row():
|
236 |
with gr.Column():
|
237 |
with gr.Row():
|
238 |
+
self.data = Data(demo)
|
239 |
with gr.Row():
|
240 |
+
self.settings = Settings()
|
|
|
241 |
with gr.Column():
|
242 |
+
self.results = Results()
|
243 |
+
self.run = gr.Button()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
244 |
|
245 |
# Update plot when dataframe is updated:
|
246 |
+
self.results.df.change(
|
247 |
plot_pareto_curve,
|
248 |
+
inputs=[self.results.df, self.settings.basic_settings.maxsize],
|
249 |
+
outputs=[self.results.pareto],
|
250 |
show_progress=False,
|
251 |
)
|
|
|
252 |
|
253 |
+
self.run.click(
|
254 |
+
create_processing_function(self, ignore=["df", "predictions_plot"]),
|
255 |
+
inputs=list(flatten_attributes(self, "interface").values()),
|
256 |
+
outputs=[self.results.df, self.results.predictions_plot],
|
257 |
+
show_progress=True,
|
258 |
+
)
|
259 |
+
|
260 |
+
|
261 |
+
def create_processing_function(interface: AppInterface, ignore=[]):
|
262 |
+
d = flatten_attributes(interface, "interface")
|
263 |
+
keys = [k.split(".")[-1] for k in d.keys()]
|
264 |
+
keys = [k for k in keys if k not in ignore]
|
265 |
+
_, idx, counts = np.unique(keys, return_index=True, return_counts=True)
|
266 |
+
if np.any(counts > 1):
|
267 |
+
raise AssertionError("Bad keys: " + ",".join(np.array(keys)[idx[counts > 1]]))
|
268 |
+
|
269 |
+
def f(components):
|
270 |
+
n = len(components)
|
271 |
+
assert n == len(keys)
|
272 |
+
return processing(**{keys[i]: components[i] for i in range(n)})
|
273 |
+
|
274 |
+
return f
|
275 |
+
|
276 |
+
|
277 |
+
class App:
|
278 |
+
def __init__(self, theme="default") -> None:
|
279 |
+
with gr.Blocks(theme=theme) as demo:
|
280 |
+
self.interface = AppInterface(demo)
|
281 |
+
|
282 |
+
demo.launch(debug=True)
|
283 |
|
284 |
|
285 |
if __name__ == "__main__":
|
286 |
+
app = App()
|
gui/data.py
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
import numpy as np
|
2 |
import pandas as pd
|
3 |
|
4 |
-
|
5 |
|
6 |
|
7 |
def generate_data(s: str, num_points: int, noise_level: float, data_seed: int):
|
|
|
1 |
import numpy as np
|
2 |
import pandas as pd
|
3 |
|
4 |
+
TEST_EQUATIONS = ["sin(2*x)/x + 0.1*x"]
|
5 |
|
6 |
|
7 |
def generate_data(s: str, num_points: int, noise_level: float, data_seed: int):
|
gui/processing.py
CHANGED
@@ -92,6 +92,7 @@ PERSISTENT_READER = None
|
|
92 |
|
93 |
|
94 |
def processing(
|
|
|
95 |
file_input,
|
96 |
force_run,
|
97 |
test_equation,
|
@@ -113,6 +114,7 @@ def processing(
|
|
113 |
optimizer_iterations,
|
114 |
batching,
|
115 |
batch_size,
|
|
|
116 |
):
|
117 |
"""Load data, then spawn a process to run the greet function."""
|
118 |
global PERSISTENT_WRITER
|
|
|
92 |
|
93 |
|
94 |
def processing(
|
95 |
+
*,
|
96 |
file_input,
|
97 |
force_run,
|
98 |
test_equation,
|
|
|
114 |
optimizer_iterations,
|
115 |
batching,
|
116 |
batch_size,
|
117 |
+
**kwargs,
|
118 |
):
|
119 |
"""Load data, then spawn a process to run the greet function."""
|
120 |
global PERSISTENT_WRITER
|