Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -18,65 +18,64 @@ if api_key:
|
|
| 18 |
client = Groq(api_key=api_key)
|
| 19 |
|
| 20 |
# ==========================================
|
| 21 |
-
# 📡 2.
|
| 22 |
# ==========================================
|
| 23 |
-
|
| 24 |
|
| 25 |
-
#
|
| 26 |
-
|
| 27 |
"User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 "
|
| 28 |
"(KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36",
|
| 29 |
"Accept": "application/json",
|
| 30 |
"Accept-Language": "en-US,en;q=0.9",
|
| 31 |
}
|
| 32 |
|
| 33 |
-
def
|
| 34 |
-
"""
|
| 35 |
for attempt in range(3):
|
| 36 |
try:
|
| 37 |
r = requests.get(
|
| 38 |
-
f"{
|
| 39 |
-
params={"
|
| 40 |
-
headers=
|
| 41 |
timeout=15
|
| 42 |
)
|
| 43 |
r.raise_for_status()
|
| 44 |
-
d = r.json()
|
| 45 |
return {
|
| 46 |
-
"last": float(d["
|
| 47 |
-
"open24h": float(d["
|
| 48 |
-
"high24h": float(d["
|
| 49 |
-
"low24h": float(d["
|
| 50 |
-
"vol24h": float(d["
|
| 51 |
-
"change_pct":
|
| 52 |
}
|
| 53 |
except Exception as e:
|
| 54 |
-
print(f"
|
| 55 |
if attempt < 2:
|
| 56 |
time.sleep(2)
|
| 57 |
return None
|
| 58 |
|
| 59 |
-
def
|
| 60 |
-
"""
|
| 61 |
for attempt in range(3):
|
| 62 |
try:
|
| 63 |
r = requests.get(
|
| 64 |
-
f"{
|
| 65 |
-
params={"
|
| 66 |
-
headers=
|
| 67 |
timeout=15
|
| 68 |
)
|
| 69 |
r.raise_for_status()
|
| 70 |
-
raw = r.json()
|
| 71 |
-
|
| 72 |
-
df = pd.DataFrame(raw, columns=
|
| 73 |
-
for
|
| 74 |
-
df[
|
| 75 |
df["ts"] = pd.to_datetime(df["ts"].astype(int), unit="ms")
|
| 76 |
-
|
| 77 |
-
return df
|
| 78 |
except Exception as e:
|
| 79 |
-
print(f"
|
| 80 |
if attempt < 2:
|
| 81 |
time.sleep(2)
|
| 82 |
return None
|
|
@@ -91,20 +90,19 @@ def compute_rsi(series, period=14):
|
|
| 91 |
rs = gain / loss
|
| 92 |
return 100 - (100 / (1 + rs))
|
| 93 |
|
| 94 |
-
def
|
| 95 |
-
"""
|
| 96 |
-
df =
|
| 97 |
if df is None or len(df) < 50:
|
| 98 |
return None, None, None, None
|
| 99 |
|
| 100 |
-
# Teknik göstergeler
|
| 101 |
df["ma_5"] = df["close"].rolling(5).mean()
|
| 102 |
df["ma_20"] = df["close"].rolling(20).mean()
|
| 103 |
df["rsi"] = compute_rsi(df["close"], 14)
|
| 104 |
df["macd"] = df["close"].ewm(span=12).mean() - df["close"].ewm(span=26).mean()
|
| 105 |
df["target"] = (df["close"].shift(-1) > df["close"]).astype(int)
|
| 106 |
-
|
| 107 |
df_model = df.dropna().reset_index(drop=True)
|
|
|
|
| 108 |
if len(df_model) < 20:
|
| 109 |
return None, None, None, None
|
| 110 |
|
|
@@ -112,14 +110,12 @@ def train_models_from_okx():
|
|
| 112 |
X = df_model[features].values
|
| 113 |
y = df_model["target"].values
|
| 114 |
|
| 115 |
-
# XGBoost
|
| 116 |
xgb_model = xgb.XGBClassifier(
|
| 117 |
n_estimators=50, max_depth=3, learning_rate=0.1,
|
| 118 |
objective="binary:logistic", eval_metric="logloss", random_state=42
|
| 119 |
)
|
| 120 |
xgb_model.fit(X, y)
|
| 121 |
|
| 122 |
-
# KNN (ölçeklendirme gerekli)
|
| 123 |
scaler = StandardScaler()
|
| 124 |
X_scaled = scaler.fit_transform(X)
|
| 125 |
knn_model = KNeighborsClassifier(n_neighbors=5)
|
|
@@ -127,56 +123,47 @@ def train_models_from_okx():
|
|
| 127 |
|
| 128 |
return xgb_model, knn_model, scaler, features
|
| 129 |
|
| 130 |
-
# Başlangıçta modelleri
|
| 131 |
-
xgb_model, knn_model, scaler, feature_cols =
|
| 132 |
|
| 133 |
if xgb_model is None:
|
| 134 |
-
|
| 135 |
-
print("OKX erişilemedi, dummy modeller kullanılıyor.")
|
| 136 |
dummy_X = np.random.rand(50, 5)
|
| 137 |
dummy_y = np.random.randint(0, 2, 50)
|
| 138 |
xgb_model = xgb.XGBClassifier(n_estimators=10, max_depth=2)
|
| 139 |
xgb_model.fit(dummy_X, dummy_y)
|
| 140 |
scaler = StandardScaler()
|
| 141 |
-
dummy_X_scaled = scaler.fit_transform(dummy_X)
|
| 142 |
knn_model = KNeighborsClassifier(n_neighbors=5)
|
| 143 |
-
knn_model.fit(
|
| 144 |
-
feature_cols = ["ma_5",
|
| 145 |
|
| 146 |
# ==========================================
|
| 147 |
# 🧠 4. ANALİZ MOTORU (XGB + KNN Ensemble)
|
| 148 |
# ==========================================
|
| 149 |
def predict_and_analyze():
|
| 150 |
-
|
| 151 |
-
|
| 152 |
-
candles = fetch_okx_candles(limit=30)
|
| 153 |
|
| 154 |
if ticker is None or candles is None:
|
| 155 |
-
return "❌
|
| 156 |
|
| 157 |
last_price = ticker["last"]
|
| 158 |
change_pct = ticker["change_pct"]
|
| 159 |
vol24h = ticker["vol24h"]
|
| 160 |
|
| 161 |
-
# 2) Teknik göstergeler
|
| 162 |
close = candles["close"]
|
| 163 |
ma5 = close.rolling(5).mean().iloc[-1]
|
| 164 |
ma20 = close.rolling(20).mean().iloc[-1]
|
| 165 |
rsi = compute_rsi(close, 14).iloc[-1]
|
| 166 |
macd = (close.ewm(span=12).mean() - close.ewm(span=26).mean()).iloc[-1]
|
| 167 |
|
| 168 |
-
# 3) XGBoost ve KNN tahminleri
|
| 169 |
feat_arr = np.array([[ma5, ma20, rsi, macd, vol24h]])
|
| 170 |
|
| 171 |
xgb_prob_up = float(xgb_model.predict_proba(feat_arr)[0][1])
|
|
|
|
| 172 |
|
| 173 |
-
feat_scaled = scaler.transform(feat_arr)
|
| 174 |
-
knn_prob_up = float(knn_model.predict_proba(feat_scaled)[0][1])
|
| 175 |
-
|
| 176 |
-
# 4) Ensemble (ortalama)
|
| 177 |
ensemble_prob_up = (xgb_prob_up + knn_prob_up) / 2
|
| 178 |
ensemble_prob_down = 1 - ensemble_prob_up
|
| 179 |
-
|
| 180 |
direction = "YUKARI 📈" if ensemble_prob_up > 0.5 else "AŞAĞI 📉"
|
| 181 |
confidence = max(ensemble_prob_up, ensemble_prob_down) * 100
|
| 182 |
|
|
@@ -189,7 +176,7 @@ def predict_and_analyze():
|
|
| 189 |
prompt = f"""
|
| 190 |
Sen uzman bir kripto para analisti ve teknik analiz uzmanısın.
|
| 191 |
|
| 192 |
-
📡
|
| 193 |
- Son Fiyat: ${last_price:,.2f}
|
| 194 |
- 24s Değişim: %{change_pct:+.2f}
|
| 195 |
- 24s Hacim: {vol24h:,.0f} BTC
|
|
@@ -201,17 +188,17 @@ def predict_and_analyze():
|
|
| 201 |
- MACD: {macd:.4f}
|
| 202 |
|
| 203 |
🤖 Model Sonuçları:
|
| 204 |
-
- XGBoost Yükseliş
|
| 205 |
-
- KNN Yükseliş
|
| 206 |
-
- Ensemble
|
| 207 |
- Tahmin Yönü: {direction} (Güven: %{confidence:.1f})
|
| 208 |
|
| 209 |
-
Türkçe bir teknik analiz raporu hazırla.
|
| 210 |
-
1. 🎯 Model Yorumu (XGBoost
|
| 211 |
2. 📈 Teknik Analiz (RSI, MA, MACD; destek/direnç)
|
| 212 |
-
3. 💡 İşlem Stratejisi (Long/Short, stop-loss, take-profit
|
| 213 |
|
| 214 |
-
Yatırım tavsiyesi olmadığını vurgula.
|
| 215 |
"""
|
| 216 |
|
| 217 |
if client:
|
|
@@ -229,7 +216,7 @@ def predict_and_analyze():
|
|
| 229 |
analysis_result = "GROQ_API_KEY bulunamadı! Hugging Face Secrets'a ekleyin."
|
| 230 |
|
| 231 |
header = (
|
| 232 |
-
f"**
|
| 233 |
f"MA5: ${ma5:,.0f} | MA20: ${ma20:,.0f}\n\n"
|
| 234 |
f"**XGB:** %{xgb_prob_up*100:.1f} | **KNN:** %{knn_prob_up*100:.1f} | "
|
| 235 |
f"**Ensemble:** %{ensemble_prob_up*100:.1f}\n\n---\n\n"
|
|
@@ -242,36 +229,19 @@ def predict_and_analyze():
|
|
| 242 |
with gr.Blocks(theme=gr.themes.Soft(primary_hue="orange", secondary_hue="slate")) as demo:
|
| 243 |
gr.Markdown(
|
| 244 |
"""
|
| 245 |
-
# ₿ BTC/USD XGBoost + KNN & Groq AI Tahmin Radarı (
|
| 246 |
-
Fiyat ve mum verileri **
|
| 247 |
"""
|
| 248 |
)
|
| 249 |
-
|
| 250 |
with gr.Row():
|
| 251 |
with gr.Column():
|
| 252 |
-
analyze_btn = gr.Button("🔄
|
| 253 |
-
gr.Markdown(
|
| 254 |
-
"""
|
| 255 |
-
**Not**: OKX API'si Cloudflare koruması altındadır. Kod, tarayıcı benzeri User-Agent ve 3 deneme mekanizması içerir.
|
| 256 |
-
"""
|
| 257 |
-
)
|
| 258 |
-
|
| 259 |
with gr.Column():
|
| 260 |
prob_output = gr.Label(label="Model Tahminleri (XGB + KNN)")
|
| 261 |
-
analysis_output = gr.Markdown(value="Butona basarak
|
| 262 |
-
|
| 263 |
-
|
| 264 |
-
fn=predict_and_analyze,
|
| 265 |
-
inputs=[],
|
| 266 |
-
outputs=[analysis_output, prob_output]
|
| 267 |
-
)
|
| 268 |
-
|
| 269 |
-
gr.Markdown(
|
| 270 |
-
"""
|
| 271 |
-
---
|
| 272 |
-
⚠️ **Yatırım Tavsiyesi Değildir**: Bu uygulama eğitim amaçlıdır.
|
| 273 |
-
"""
|
| 274 |
-
)
|
| 275 |
|
| 276 |
if __name__ == "__main__":
|
| 277 |
demo.launch()
|
|
|
|
| 18 |
client = Groq(api_key=api_key)
|
| 19 |
|
| 20 |
# ==========================================
|
| 21 |
+
# 📡 2. MEXC PUBLIC API
|
| 22 |
# ==========================================
|
| 23 |
+
MEXC_BASE = "https://api.mexc.com"
|
| 24 |
|
| 25 |
+
# Akamai WAF'ın Python varsayılan UA'sını engellemesini aşmak için
|
| 26 |
+
MEXC_HEADERS = {
|
| 27 |
"User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 "
|
| 28 |
"(KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36",
|
| 29 |
"Accept": "application/json",
|
| 30 |
"Accept-Language": "en-US,en;q=0.9",
|
| 31 |
}
|
| 32 |
|
| 33 |
+
def fetch_mexc_ticker(symbol="BTCUSDT"):
|
| 34 |
+
"""MEXC spot son fiyat. 3 kez dener."""
|
| 35 |
for attempt in range(3):
|
| 36 |
try:
|
| 37 |
r = requests.get(
|
| 38 |
+
f"{MEXC_BASE}/api/v3/ticker/24hr",
|
| 39 |
+
params={"symbol": symbol},
|
| 40 |
+
headers=MEXC_HEADERS,
|
| 41 |
timeout=15
|
| 42 |
)
|
| 43 |
r.raise_for_status()
|
| 44 |
+
d = r.json()
|
| 45 |
return {
|
| 46 |
+
"last": float(d["lastPrice"]),
|
| 47 |
+
"open24h": float(d["openPrice"]),
|
| 48 |
+
"high24h": float(d["highPrice"]),
|
| 49 |
+
"low24h": float(d["lowPrice"]),
|
| 50 |
+
"vol24h": float(d["volume"]),
|
| 51 |
+
"change_pct": float(d["priceChangePercent"]),
|
| 52 |
}
|
| 53 |
except Exception as e:
|
| 54 |
+
print(f"MEXC ticker attempt {attempt+1} failed: {e}")
|
| 55 |
if attempt < 2:
|
| 56 |
time.sleep(2)
|
| 57 |
return None
|
| 58 |
|
| 59 |
+
def fetch_mexc_klines(symbol="BTCUSDT", interval="1h", limit=100):
|
| 60 |
+
"""MEXC spot K-line. 3 kez dener."""
|
| 61 |
for attempt in range(3):
|
| 62 |
try:
|
| 63 |
r = requests.get(
|
| 64 |
+
f"{MEXC_BASE}/api/v3/klines",
|
| 65 |
+
params={"symbol": symbol, "interval": interval, "limit": limit},
|
| 66 |
+
headers=MEXC_HEADERS,
|
| 67 |
timeout=15
|
| 68 |
)
|
| 69 |
r.raise_for_status()
|
| 70 |
+
raw = r.json()
|
| 71 |
+
cols = ["ts","open","high","low","close","vol","close_time","qvol","trades","tb","tq","ignore"]
|
| 72 |
+
df = pd.DataFrame(raw, columns=cols)
|
| 73 |
+
for c in ["open","high","low","close","vol"]:
|
| 74 |
+
df[c] = df[c].astype(float)
|
| 75 |
df["ts"] = pd.to_datetime(df["ts"].astype(int), unit="ms")
|
| 76 |
+
return df[["ts","open","high","low","close","vol"]].sort_values("ts").reset_index(drop=True)
|
|
|
|
| 77 |
except Exception as e:
|
| 78 |
+
print(f"MEXC klines attempt {attempt+1} failed: {e}")
|
| 79 |
if attempt < 2:
|
| 80 |
time.sleep(2)
|
| 81 |
return None
|
|
|
|
| 90 |
rs = gain / loss
|
| 91 |
return 100 - (100 / (1 + rs))
|
| 92 |
|
| 93 |
+
def train_models_from_mexc():
|
| 94 |
+
"""MEXC'ten çekilen mum verisiyle XGBoost ve KNN eğitir."""
|
| 95 |
+
df = fetch_mexc_klines(limit=500)
|
| 96 |
if df is None or len(df) < 50:
|
| 97 |
return None, None, None, None
|
| 98 |
|
|
|
|
| 99 |
df["ma_5"] = df["close"].rolling(5).mean()
|
| 100 |
df["ma_20"] = df["close"].rolling(20).mean()
|
| 101 |
df["rsi"] = compute_rsi(df["close"], 14)
|
| 102 |
df["macd"] = df["close"].ewm(span=12).mean() - df["close"].ewm(span=26).mean()
|
| 103 |
df["target"] = (df["close"].shift(-1) > df["close"]).astype(int)
|
|
|
|
| 104 |
df_model = df.dropna().reset_index(drop=True)
|
| 105 |
+
|
| 106 |
if len(df_model) < 20:
|
| 107 |
return None, None, None, None
|
| 108 |
|
|
|
|
| 110 |
X = df_model[features].values
|
| 111 |
y = df_model["target"].values
|
| 112 |
|
|
|
|
| 113 |
xgb_model = xgb.XGBClassifier(
|
| 114 |
n_estimators=50, max_depth=3, learning_rate=0.1,
|
| 115 |
objective="binary:logistic", eval_metric="logloss", random_state=42
|
| 116 |
)
|
| 117 |
xgb_model.fit(X, y)
|
| 118 |
|
|
|
|
| 119 |
scaler = StandardScaler()
|
| 120 |
X_scaled = scaler.fit_transform(X)
|
| 121 |
knn_model = KNeighborsClassifier(n_neighbors=5)
|
|
|
|
| 123 |
|
| 124 |
return xgb_model, knn_model, scaler, features
|
| 125 |
|
| 126 |
+
# Başlangıçta modelleri MEXC verisiyle eğit
|
| 127 |
+
xgb_model, knn_model, scaler, feature_cols = train_models_from_mexc()
|
| 128 |
|
| 129 |
if xgb_model is None:
|
| 130 |
+
print("MEXC erişilemedi, dummy modeller kullanılıyor.")
|
|
|
|
| 131 |
dummy_X = np.random.rand(50, 5)
|
| 132 |
dummy_y = np.random.randint(0, 2, 50)
|
| 133 |
xgb_model = xgb.XGBClassifier(n_estimators=10, max_depth=2)
|
| 134 |
xgb_model.fit(dummy_X, dummy_y)
|
| 135 |
scaler = StandardScaler()
|
|
|
|
| 136 |
knn_model = KNeighborsClassifier(n_neighbors=5)
|
| 137 |
+
knn_model.fit(scaler.fit_transform(dummy_X), dummy_y)
|
| 138 |
+
feature_cols = ["ma_5","ma_20","rsi","macd","vol"]
|
| 139 |
|
| 140 |
# ==========================================
|
| 141 |
# 🧠 4. ANALİZ MOTORU (XGB + KNN Ensemble)
|
| 142 |
# ==========================================
|
| 143 |
def predict_and_analyze():
|
| 144 |
+
ticker = fetch_mexc_ticker()
|
| 145 |
+
candles = fetch_mexc_klines(limit=30)
|
|
|
|
| 146 |
|
| 147 |
if ticker is None or candles is None:
|
| 148 |
+
return "❌ MEXC API'sine erişilemedi (Akamai WAF veya Hugging Face ağ kısıtlaması). Lütfen sonra tekrar deneyin.", {}
|
| 149 |
|
| 150 |
last_price = ticker["last"]
|
| 151 |
change_pct = ticker["change_pct"]
|
| 152 |
vol24h = ticker["vol24h"]
|
| 153 |
|
|
|
|
| 154 |
close = candles["close"]
|
| 155 |
ma5 = close.rolling(5).mean().iloc[-1]
|
| 156 |
ma20 = close.rolling(20).mean().iloc[-1]
|
| 157 |
rsi = compute_rsi(close, 14).iloc[-1]
|
| 158 |
macd = (close.ewm(span=12).mean() - close.ewm(span=26).mean()).iloc[-1]
|
| 159 |
|
|
|
|
| 160 |
feat_arr = np.array([[ma5, ma20, rsi, macd, vol24h]])
|
| 161 |
|
| 162 |
xgb_prob_up = float(xgb_model.predict_proba(feat_arr)[0][1])
|
| 163 |
+
knn_prob_up = float(knn_model.predict_proba(scaler.transform(feat_arr))[0][1])
|
| 164 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 165 |
ensemble_prob_up = (xgb_prob_up + knn_prob_up) / 2
|
| 166 |
ensemble_prob_down = 1 - ensemble_prob_up
|
|
|
|
| 167 |
direction = "YUKARI 📈" if ensemble_prob_up > 0.5 else "AŞAĞI 📉"
|
| 168 |
confidence = max(ensemble_prob_up, ensemble_prob_down) * 100
|
| 169 |
|
|
|
|
| 176 |
prompt = f"""
|
| 177 |
Sen uzman bir kripto para analisti ve teknik analiz uzmanısın.
|
| 178 |
|
| 179 |
+
📡 MEXC Canlı Verileri (BTC/USDT):
|
| 180 |
- Son Fiyat: ${last_price:,.2f}
|
| 181 |
- 24s Değişim: %{change_pct:+.2f}
|
| 182 |
- 24s Hacim: {vol24h:,.0f} BTC
|
|
|
|
| 188 |
- MACD: {macd:.4f}
|
| 189 |
|
| 190 |
🤖 Model Sonuçları:
|
| 191 |
+
- XGBoost Yükseliş: %{xgb_prob_up*100:.1f}
|
| 192 |
+
- KNN Yükseliş: %{knn_prob_up*100:.1f}
|
| 193 |
+
- Ensemble Yükseliş: %{ensemble_prob_up*100:.1f}
|
| 194 |
- Tahmin Yönü: {direction} (Güven: %{confidence:.1f})
|
| 195 |
|
| 196 |
+
Türkçe bir teknik analiz raporu hazırla. Başlıklar:
|
| 197 |
+
1. 🎯 Model Yorumu (XGBoost + KNN karşılaştırması)
|
| 198 |
2. 📈 Teknik Analiz (RSI, MA, MACD; destek/direnç)
|
| 199 |
+
3. 💡 İşlem Stratejisi (Long/Short, stop-loss, take-profit)
|
| 200 |
|
| 201 |
+
Yatırım tavsiyesi olmadığını vurgula.
|
| 202 |
"""
|
| 203 |
|
| 204 |
if client:
|
|
|
|
| 216 |
analysis_result = "GROQ_API_KEY bulunamadı! Hugging Face Secrets'a ekleyin."
|
| 217 |
|
| 218 |
header = (
|
| 219 |
+
f"**MEXC Canlı: ${last_price:,.2f}** (%{change_pct:+.2f}) | RSI: {rsi:.1f} | "
|
| 220 |
f"MA5: ${ma5:,.0f} | MA20: ${ma20:,.0f}\n\n"
|
| 221 |
f"**XGB:** %{xgb_prob_up*100:.1f} | **KNN:** %{knn_prob_up*100:.1f} | "
|
| 222 |
f"**Ensemble:** %{ensemble_prob_up*100:.1f}\n\n---\n\n"
|
|
|
|
| 229 |
with gr.Blocks(theme=gr.themes.Soft(primary_hue="orange", secondary_hue="slate")) as demo:
|
| 230 |
gr.Markdown(
|
| 231 |
"""
|
| 232 |
+
# ₿ BTC/USD XGBoost + KNN & Groq AI Tahmin Radarı (MEXC Canlı)
|
| 233 |
+
Fiyat ve mum verileri **MEXC halka açık API**'sinden otomatik çekilir. **XGBoost + KNN** ensemble tahmini yapar, **Groq** analiz raporu üretir.
|
| 234 |
"""
|
| 235 |
)
|
|
|
|
| 236 |
with gr.Row():
|
| 237 |
with gr.Column():
|
| 238 |
+
analyze_btn = gr.Button("🔄 MEXC'ten Veri Çek ve Analiz Et", variant="primary", size="lg")
|
| 239 |
+
gr.Markdown("**Not**: MEXC API'si Akamai WAF koruması altındadır; kod tarayıcı UA ve 3 deneme içerir.")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 240 |
with gr.Column():
|
| 241 |
prob_output = gr.Label(label="Model Tahminleri (XGB + KNN)")
|
| 242 |
+
analysis_output = gr.Markdown(value="Butona basarak MEXC'ten canlı veri çekip analizi başlatın.")
|
| 243 |
+
analyze_btn.click(fn=predict_and_analyze, inputs=[], outputs=[analysis_output, prob_output])
|
| 244 |
+
gr.Markdown("---\n⚠️ **Yatırım Tavsiyesi Değildir**: Bu uygulama eğitim amaçlıdır.")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 245 |
|
| 246 |
if __name__ == "__main__":
|
| 247 |
demo.launch()
|