Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,103 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import pandas as pd
|
3 |
+
import os
|
4 |
+
import json
|
5 |
+
import requests
|
6 |
+
from ctgan import CTGAN
|
7 |
+
from sklearn.preprocessing import LabelEncoder
|
8 |
+
|
9 |
+
def generate_schema(prompt):
|
10 |
+
"""Fetches schema from Hugging Face Spaces API."""
|
11 |
+
API_URL = "https://infinitymatter-Synthetic_Data_Generator_SRIJAN.hf.space/run/predict"
|
12 |
+
|
13 |
+
# Fetch API token securely
|
14 |
+
hf_token = st.secrets["hf_token"]
|
15 |
+
headers = {"Authorization": f"Bearer {hf_token}"}
|
16 |
+
|
17 |
+
payload = {"data": [prompt]}
|
18 |
+
|
19 |
+
try:
|
20 |
+
response = requests.post(API_URL, headers=headers, json=payload)
|
21 |
+
response.raise_for_status()
|
22 |
+
schema = response.json()
|
23 |
+
|
24 |
+
if 'columns' not in schema or 'types' not in schema or 'size' not in schema:
|
25 |
+
raise ValueError("Invalid schema format!")
|
26 |
+
|
27 |
+
return schema
|
28 |
+
except requests.exceptions.RequestException as e:
|
29 |
+
st.error(f"β API request failed: {e}")
|
30 |
+
return None
|
31 |
+
except json.JSONDecodeError:
|
32 |
+
st.error("β Failed to parse JSON response.")
|
33 |
+
return None
|
34 |
+
|
35 |
+
|
36 |
+
def train_and_generate_synthetic(real_data, schema, output_path):
|
37 |
+
"""Trains a CTGAN model and generates synthetic data."""
|
38 |
+
categorical_cols = [col for col, dtype in zip(schema['columns'], schema['types']) if dtype == 'string']
|
39 |
+
|
40 |
+
# Store label encoders
|
41 |
+
label_encoders = {}
|
42 |
+
for col in categorical_cols:
|
43 |
+
le = LabelEncoder()
|
44 |
+
real_data[col] = le.fit_transform(real_data[col])
|
45 |
+
label_encoders[col] = le
|
46 |
+
|
47 |
+
# Train CTGAN
|
48 |
+
gan = CTGAN(epochs=300)
|
49 |
+
gan.fit(real_data, categorical_cols)
|
50 |
+
|
51 |
+
# Generate synthetic data
|
52 |
+
synthetic_data = gan.sample(schema['size'])
|
53 |
+
|
54 |
+
# Decode categorical columns
|
55 |
+
for col in categorical_cols:
|
56 |
+
synthetic_data[col] = label_encoders[col].inverse_transform(synthetic_data[col])
|
57 |
+
|
58 |
+
# Save to CSV
|
59 |
+
os.makedirs('outputs', exist_ok=True)
|
60 |
+
synthetic_data.to_csv(output_path, index=False)
|
61 |
+
st.success(f"β
Synthetic data saved to {output_path}")
|
62 |
+
|
63 |
+
def fetch_data(domain):
|
64 |
+
"""Fetches real data for the given domain and ensures it's a valid DataFrame."""
|
65 |
+
data_path = f"datasets/{domain}.csv"
|
66 |
+
if os.path.exists(data_path):
|
67 |
+
df = pd.read_csv(data_path)
|
68 |
+
if not isinstance(df, pd.DataFrame) or df.empty:
|
69 |
+
raise ValueError("β Loaded data is invalid!")
|
70 |
+
return df
|
71 |
+
else:
|
72 |
+
st.error(f"β Dataset for {domain} not found.")
|
73 |
+
return None
|
74 |
+
|
75 |
+
st.title("β¨ AI-Powered Synthetic Dataset Generator")
|
76 |
+
st.write("Give a short description of the dataset you need, and AI will generate it for you using real data + GANs!")
|
77 |
+
|
78 |
+
# User input
|
79 |
+
user_prompt = st.text_input("Describe the dataset (e.g., 'Create dataset for hospital patients')", "")
|
80 |
+
domain = st.selectbox("Select Domain for Real Data", ["healthcare", "finance", "retail", "other"])
|
81 |
+
|
82 |
+
data = None
|
83 |
+
if st.button("Generate Schema"):
|
84 |
+
if user_prompt.strip():
|
85 |
+
with st.spinner("Generating schema..."):
|
86 |
+
schema = generate_schema(user_prompt)
|
87 |
+
|
88 |
+
if schema is None:
|
89 |
+
st.error("β Schema generation failed. Please check API response.")
|
90 |
+
else:
|
91 |
+
st.success("β
Schema generated successfully!")
|
92 |
+
st.json(schema)
|
93 |
+
data = fetch_data(domain)
|
94 |
+
else:
|
95 |
+
st.warning("β οΈ Please enter a dataset description before generating the schema.")
|
96 |
+
|
97 |
+
if data is not None and schema is not None:
|
98 |
+
output_path = "outputs/synthetic_data.csv"
|
99 |
+
if st.button("Generate Synthetic Data"):
|
100 |
+
with st.spinner("Training GAN and generating synthetic data..."):
|
101 |
+
train_and_generate_synthetic(data, schema, output_path)
|
102 |
+
with open(output_path, "rb") as file:
|
103 |
+
st.download_button("Download Synthetic Data", file, file_name="synthetic_data.csv", mime="text/csv")
|