try new dataset handling
Browse files
app.py
CHANGED
@@ -1,79 +1,102 @@
|
|
|
|
|
|
1 |
import gradio as gr
|
2 |
import numpy as np
|
|
|
3 |
import matplotlib
|
4 |
import matplotlib.pyplot as plt
|
5 |
-
import pandas as pd
|
6 |
|
7 |
from datasets import load_dataset
|
8 |
from sklearn.ensemble import GradientBoostingClassifier
|
9 |
from sklearn.model_selection import train_test_split
|
10 |
from sklearn.metrics import accuracy_score, confusion_matrix
|
11 |
|
12 |
-
|
|
|
13 |
|
14 |
-
|
15 |
-
#
|
|
|
|
|
|
|
|
|
16 |
SUGGESTED_DATASETS = [
|
17 |
-
"
|
18 |
-
"
|
19 |
-
"SKIP/ENTER_CUSTOM"
|
20 |
]
|
21 |
|
22 |
-
|
|
|
23 |
"""
|
24 |
-
Loads
|
25 |
-
|
26 |
-
returns X, y as NumPy arrays for modeling.
|
27 |
"""
|
28 |
-
#
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
# Subset to selected columns
|
36 |
-
if label_column not in df.columns:
|
37 |
-
raise ValueError(f"Label column '{label_column}' not in dataset columns: {df.columns.to_list()}")
|
38 |
-
|
39 |
-
for col in feature_columns:
|
40 |
-
if col not in df.columns:
|
41 |
-
raise ValueError(f"Feature column '{col}' not in dataset columns: {df.columns.to_list()}")
|
42 |
-
|
43 |
-
# Split into X and y
|
44 |
-
X = df[feature_columns].values
|
45 |
-
y = df[label_column].values
|
46 |
-
|
47 |
-
return X, y, df.columns.tolist()
|
48 |
|
49 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
50 |
learning_rate, n_estimators, max_depth, test_size):
|
51 |
"""
|
52 |
-
1. Determine final dataset ID (
|
53 |
-
2. Load dataset ->
|
54 |
-
3. Train
|
55 |
-
4.
|
|
|
|
|
56 |
"""
|
57 |
-
|
58 |
-
# Decide which dataset ID to use
|
59 |
if dataset_id != "SKIP/ENTER_CUSTOM":
|
60 |
final_id = dataset_id
|
61 |
else:
|
62 |
-
# Use the user-supplied "custom_dataset_id"
|
63 |
final_id = custom_dataset_id.strip()
|
64 |
|
65 |
-
#
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
73 |
X_train, X_test, y_train, y_test = train_test_split(
|
74 |
X, y, test_size=test_size, random_state=42
|
75 |
)
|
76 |
-
|
77 |
# Train model
|
78 |
clf = GradientBoostingClassifier(
|
79 |
learning_rate=learning_rate,
|
@@ -82,13 +105,15 @@ def train_model(dataset_id, custom_dataset_id, label_column, feature_columns,
|
|
82 |
random_state=42
|
83 |
)
|
84 |
clf.fit(X_train, y_train)
|
85 |
-
|
86 |
-
#
|
87 |
y_pred = clf.predict(X_test)
|
88 |
accuracy = accuracy_score(y_test, y_pred)
|
89 |
cm = confusion_matrix(y_test, y_pred)
|
90 |
|
91 |
-
#
|
|
|
|
|
92 |
fig, axs = plt.subplots(1, 2, figsize=(10, 4))
|
93 |
|
94 |
# Subplot 1: Feature Importances
|
@@ -103,90 +128,80 @@ def train_model(dataset_id, custom_dataset_id, label_column, feature_columns,
|
|
103 |
im = axs[1].imshow(cm, interpolation='nearest', cmap=plt.cm.Blues)
|
104 |
axs[1].set_title("Confusion Matrix")
|
105 |
plt.colorbar(im, ax=axs[1])
|
106 |
-
# Labeling
|
107 |
axs[1].set_xlabel("Predicted")
|
108 |
axs[1].set_ylabel("True")
|
109 |
|
110 |
-
#
|
111 |
thresh = cm.max() / 2.0
|
112 |
for i in range(cm.shape[0]):
|
113 |
for j in range(cm.shape[1]):
|
114 |
color = "white" if cm[i, j] > thresh else "black"
|
115 |
-
axs[1].text(j, i,
|
116 |
|
117 |
plt.tight_layout()
|
118 |
|
119 |
-
|
120 |
-
|
121 |
-
|
|
|
|
|
|
|
|
|
122 |
|
123 |
-
return
|
124 |
|
125 |
-
def update_columns(dataset_id, dataset_config, custom_dataset_id):
|
126 |
-
"""
|
127 |
-
Load the dataset from HF hub, using either the suggested one or the custom user-specified,
|
128 |
-
plus an optional config.
|
129 |
-
"""
|
130 |
-
if dataset_id != "SKIP/ENTER_CUSTOM":
|
131 |
-
final_id = dataset_id
|
132 |
-
final_config = dataset_config.strip() if dataset_config else None
|
133 |
-
else:
|
134 |
-
# Use the user-supplied text
|
135 |
-
final_id = custom_dataset_id.strip()
|
136 |
-
final_config = None # or parse from text if you like
|
137 |
-
|
138 |
-
try:
|
139 |
-
if final_config:
|
140 |
-
ds = load_dataset(final_id, final_config, split="train")
|
141 |
-
else:
|
142 |
-
ds = load_dataset(final_id, split="train")
|
143 |
-
df = pd.DataFrame(ds)
|
144 |
-
cols = df.columns.tolist()
|
145 |
-
return gr.update(choices=cols), gr.update(choices=cols), f"Columns found: {cols}"
|
146 |
-
except Exception as e:
|
147 |
-
return gr.update(choices=[]), gr.update(choices=[]), f"Error loading {final_id}: {e}"
|
148 |
|
|
|
149 |
with gr.Blocks() as demo:
|
150 |
-
gr.Markdown("
|
151 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
152 |
with gr.Row():
|
153 |
dataset_dropdown = gr.Dropdown(
|
|
|
154 |
choices=SUGGESTED_DATASETS,
|
155 |
-
value=SUGGESTED_DATASETS[0]
|
156 |
-
|
|
|
|
|
|
|
157 |
)
|
158 |
-
|
159 |
-
|
160 |
-
|
161 |
-
# Button to load columns from the chosen dataset
|
162 |
-
load_cols_btn = gr.Button("Load columns")
|
163 |
load_cols_info = gr.Markdown()
|
164 |
-
|
|
|
165 |
with gr.Row():
|
166 |
label_col = gr.Dropdown(choices=[], label="Label column (choose 1)")
|
167 |
feature_cols = gr.CheckboxGroup(choices=[], label="Feature columns (choose 1 or more)")
|
168 |
|
169 |
-
#
|
170 |
learning_rate_slider = gr.Slider(0.01, 1.0, value=0.1, step=0.01, label="learning_rate")
|
171 |
n_estimators_slider = gr.Slider(50, 300, value=100, step=50, label="n_estimators")
|
172 |
max_depth_slider = gr.Slider(1, 10, value=3, step=1, label="max_depth")
|
173 |
-
test_size_slider = gr.Slider(0.1, 0.9, value=0.3, step=0.1, label="test_size (
|
174 |
|
175 |
train_button = gr.Button("Train & Evaluate")
|
176 |
-
|
177 |
output_text = gr.Markdown()
|
178 |
output_plot = gr.Plot()
|
179 |
-
# We might also want to show the columns for reference post-training
|
180 |
-
columns_return = gr.Markdown()
|
181 |
|
182 |
-
#
|
183 |
load_cols_btn.click(
|
184 |
fn=update_columns,
|
185 |
inputs=[dataset_dropdown, custom_dataset_id],
|
186 |
-
outputs=[label_col, feature_cols, load_cols_info]
|
187 |
)
|
188 |
|
189 |
-
#
|
190 |
train_button.click(
|
191 |
fn=train_model,
|
192 |
inputs=[
|
@@ -199,7 +214,7 @@ with gr.Blocks() as demo:
|
|
199 |
max_depth_slider,
|
200 |
test_size_slider
|
201 |
],
|
202 |
-
outputs=[output_text, output_plot,
|
203 |
)
|
204 |
|
205 |
demo.launch()
|
|
|
1 |
+
# app.py
|
2 |
+
|
3 |
import gradio as gr
|
4 |
import numpy as np
|
5 |
+
import pandas as pd
|
6 |
import matplotlib
|
7 |
import matplotlib.pyplot as plt
|
|
|
8 |
|
9 |
from datasets import load_dataset
|
10 |
from sklearn.ensemble import GradientBoostingClassifier
|
11 |
from sklearn.model_selection import train_test_split
|
12 |
from sklearn.metrics import accuracy_score, confusion_matrix
|
13 |
|
14 |
+
# In some remote environments, Matplotlib needs to be set to 'Agg' backend
|
15 |
+
matplotlib.use('Agg')
|
16 |
|
17 |
+
################################################################################
|
18 |
+
# SUGGESTED_DATASETS: Must actually exist on huggingface.co/datasets.
|
19 |
+
#
|
20 |
+
# "scikit-learn/iris" -> a tabular Iris dataset with a "train" split of 150 rows.
|
21 |
+
# "uci/wine" -> a tabular Wine dataset with a "train" split of 178 rows.
|
22 |
+
################################################################################
|
23 |
SUGGESTED_DATASETS = [
|
24 |
+
"scikit-learn/iris",
|
25 |
+
"uci/wine",
|
26 |
+
"SKIP/ENTER_CUSTOM" # a placeholder meaning "use custom_dataset_id"
|
27 |
]
|
28 |
|
29 |
+
|
30 |
+
def update_columns(dataset_id, custom_dataset_id):
|
31 |
"""
|
32 |
+
Loads the chosen dataset (train split) and returns its column names,
|
33 |
+
to populate the Label Column & Feature Columns selectors.
|
|
|
34 |
"""
|
35 |
+
# If user picked a suggested dataset (not SKIP), use that
|
36 |
+
if dataset_id != "SKIP/ENTER_CUSTOM":
|
37 |
+
final_id = dataset_id
|
38 |
+
else:
|
39 |
+
# Use the user-supplied dataset ID
|
40 |
+
final_id = custom_dataset_id.strip()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
41 |
|
42 |
+
try:
|
43 |
+
# Load just the "train" split; many HF datasets have train/test/validation
|
44 |
+
ds = load_dataset(final_id, split="train")
|
45 |
+
df = pd.DataFrame(ds)
|
46 |
+
cols = df.columns.tolist()
|
47 |
+
|
48 |
+
message = f"**Loaded dataset**: {final_id}\n\n**Columns found**: {cols}"
|
49 |
+
# Return list of columns for both label & features
|
50 |
+
return (
|
51 |
+
gr.update(choices=cols, value=None), # label_col dropdown
|
52 |
+
gr.update(choices=cols, value=[]), # feature_cols checkbox group
|
53 |
+
message
|
54 |
+
)
|
55 |
+
except Exception as e:
|
56 |
+
# If load fails or dataset doesn't exist
|
57 |
+
err_msg = f"**Error loading** `{final_id}`: {e}"
|
58 |
+
return (
|
59 |
+
gr.update(choices=[], value=None),
|
60 |
+
gr.update(choices=[], value=[]),
|
61 |
+
err_msg
|
62 |
+
)
|
63 |
+
|
64 |
+
|
65 |
+
def train_model(dataset_id, custom_dataset_id, label_column, feature_columns,
|
66 |
learning_rate, n_estimators, max_depth, test_size):
|
67 |
"""
|
68 |
+
1. Determine the final dataset ID (from dropdown or custom text).
|
69 |
+
2. Load the dataset -> create dataframe -> X, y.
|
70 |
+
3. Train GradientBoostingClassifier.
|
71 |
+
4. Return metrics (accuracy) and a Matplotlib figure with:
|
72 |
+
- Feature importance bar chart
|
73 |
+
- Confusion matrix heatmap
|
74 |
"""
|
|
|
|
|
75 |
if dataset_id != "SKIP/ENTER_CUSTOM":
|
76 |
final_id = dataset_id
|
77 |
else:
|
|
|
78 |
final_id = custom_dataset_id.strip()
|
79 |
|
80 |
+
# Load dataset
|
81 |
+
ds = load_dataset(final_id, split="train")
|
82 |
+
df = pd.DataFrame(ds)
|
83 |
+
|
84 |
+
# Basic validation
|
85 |
+
if label_column not in df.columns:
|
86 |
+
raise ValueError(f"Label column '{label_column}' not found in dataset columns.")
|
87 |
+
for fc in feature_columns:
|
88 |
+
if fc not in df.columns:
|
89 |
+
raise ValueError(f"Feature column '{fc}' not found in dataset columns.")
|
90 |
+
|
91 |
+
# Build X, y arrays
|
92 |
+
X = df[feature_columns].values
|
93 |
+
y = df[label_column].values
|
94 |
+
|
95 |
+
# Split
|
96 |
X_train, X_test, y_train, y_test = train_test_split(
|
97 |
X, y, test_size=test_size, random_state=42
|
98 |
)
|
99 |
+
|
100 |
# Train model
|
101 |
clf = GradientBoostingClassifier(
|
102 |
learning_rate=learning_rate,
|
|
|
105 |
random_state=42
|
106 |
)
|
107 |
clf.fit(X_train, y_train)
|
108 |
+
|
109 |
+
# Predictions & metrics
|
110 |
y_pred = clf.predict(X_test)
|
111 |
accuracy = accuracy_score(y_test, y_pred)
|
112 |
cm = confusion_matrix(y_test, y_pred)
|
113 |
|
114 |
+
# Build a single figure with 2 subplots:
|
115 |
+
# 1) Feature importances
|
116 |
+
# 2) Confusion matrix heatmap
|
117 |
fig, axs = plt.subplots(1, 2, figsize=(10, 4))
|
118 |
|
119 |
# Subplot 1: Feature Importances
|
|
|
128 |
im = axs[1].imshow(cm, interpolation='nearest', cmap=plt.cm.Blues)
|
129 |
axs[1].set_title("Confusion Matrix")
|
130 |
plt.colorbar(im, ax=axs[1])
|
|
|
131 |
axs[1].set_xlabel("Predicted")
|
132 |
axs[1].set_ylabel("True")
|
133 |
|
134 |
+
# Optionally annotate each cell with the count
|
135 |
thresh = cm.max() / 2.0
|
136 |
for i in range(cm.shape[0]):
|
137 |
for j in range(cm.shape[1]):
|
138 |
color = "white" if cm[i, j] > thresh else "black"
|
139 |
+
axs[1].text(j, i, str(cm[i, j]), ha="center", va="center", color=color)
|
140 |
|
141 |
plt.tight_layout()
|
142 |
|
143 |
+
# Build textual summary
|
144 |
+
text_summary = (
|
145 |
+
f"**Dataset used**: `{final_id}`\n\n"
|
146 |
+
f"**Label column**: `{label_column}`\n\n"
|
147 |
+
f"**Feature columns**: `{feature_columns}`\n\n"
|
148 |
+
f"**Accuracy**: {accuracy:.3f}\n\n"
|
149 |
+
)
|
150 |
|
151 |
+
return text_summary, fig
|
152 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
153 |
|
154 |
+
# Build the Gradio Blocks UI
|
155 |
with gr.Blocks() as demo:
|
156 |
+
gr.Markdown("# Train a GradientBoostingClassifier on any HF Dataset\n")
|
157 |
+
gr.Markdown(
|
158 |
+
"1. Choose a suggested dataset from the dropdown **or** enter a custom dataset ID in the format `user/dataset`.\n"
|
159 |
+
"2. Click **Load Columns** to inspect the columns.\n"
|
160 |
+
"3. Pick a **Label column** and **Feature columns**.\n"
|
161 |
+
"4. Adjust hyperparameters and click **Train & Evaluate**.\n"
|
162 |
+
"5. Observe accuracy, feature importances, and a confusion matrix heatmap.\n\n"
|
163 |
+
"*(Note: the dataset must have a `train` split!)*"
|
164 |
+
)
|
165 |
+
|
166 |
+
# Row 1: Dataset selection
|
167 |
with gr.Row():
|
168 |
dataset_dropdown = gr.Dropdown(
|
169 |
+
label="Choose suggested dataset",
|
170 |
choices=SUGGESTED_DATASETS,
|
171 |
+
value=SUGGESTED_DATASETS[0] # default
|
172 |
+
)
|
173 |
+
custom_dataset_id = gr.Textbox(
|
174 |
+
label="Or enter a custom dataset ID",
|
175 |
+
placeholder="e.g. username/my_custom_dataset"
|
176 |
)
|
177 |
+
|
178 |
+
load_cols_btn = gr.Button("Load Columns")
|
|
|
|
|
|
|
179 |
load_cols_info = gr.Markdown()
|
180 |
+
|
181 |
+
# Row 2: label & feature columns
|
182 |
with gr.Row():
|
183 |
label_col = gr.Dropdown(choices=[], label="Label column (choose 1)")
|
184 |
feature_cols = gr.CheckboxGroup(choices=[], label="Feature columns (choose 1 or more)")
|
185 |
|
186 |
+
# Hyperparameters
|
187 |
learning_rate_slider = gr.Slider(0.01, 1.0, value=0.1, step=0.01, label="learning_rate")
|
188 |
n_estimators_slider = gr.Slider(50, 300, value=100, step=50, label="n_estimators")
|
189 |
max_depth_slider = gr.Slider(1, 10, value=3, step=1, label="max_depth")
|
190 |
+
test_size_slider = gr.Slider(0.1, 0.9, value=0.3, step=0.1, label="test_size fraction (0.1-0.9)")
|
191 |
|
192 |
train_button = gr.Button("Train & Evaluate")
|
193 |
+
|
194 |
output_text = gr.Markdown()
|
195 |
output_plot = gr.Plot()
|
|
|
|
|
196 |
|
197 |
+
# Link the "Load Columns" button -> update_columns function
|
198 |
load_cols_btn.click(
|
199 |
fn=update_columns,
|
200 |
inputs=[dataset_dropdown, custom_dataset_id],
|
201 |
+
outputs=[label_col, feature_cols, load_cols_info],
|
202 |
)
|
203 |
|
204 |
+
# Link "Train & Evaluate" -> train_model function
|
205 |
train_button.click(
|
206 |
fn=train_model,
|
207 |
inputs=[
|
|
|
214 |
max_depth_slider,
|
215 |
test_size_slider
|
216 |
],
|
217 |
+
outputs=[output_text, output_plot],
|
218 |
)
|
219 |
|
220 |
demo.launch()
|