daafa999 commited on
Commit
a8832e3
Β·
verified Β·
1 Parent(s): 3635c9e

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +26 -38
app.py CHANGED
@@ -1,5 +1,4 @@
1
  import gradio as gr
2
- from fastapi import FastAPI, HTTPException
3
  import ccxt
4
  import pandas as pd
5
  import pandas_ta as ta
@@ -9,13 +8,10 @@ from sklearn.preprocessing import MinMaxScaler
9
  import joblib
10
  import os
11
  import warnings
12
- import uvicorn
13
 
14
  warnings.filterwarnings('ignore')
15
 
16
- # πŸ”Ή 1. FastAPI App
17
- fastapi_app = FastAPI(title="Crypto ML Strategy (Paper 2105.06827v3)")
18
-
19
  SYMBOL = "ETH/USDT"
20
  TIMEFRAME = "4h"
21
  FEE = 0.0015
@@ -37,6 +33,7 @@ def prepare_dataset(df: pd.DataFrame):
37
  df.ta.bbands(append=True)
38
  df.ta.cci(append=True)
39
  df.dropna(inplace=True)
 
40
  df["next_close"] = df["close"].shift(-1)
41
  df["label"] = (df["next_close"] > df["close"] * (1 + FEE)).astype(int)
42
  df.dropna(inplace=True)
@@ -54,8 +51,7 @@ def prepare_dataset(df: pd.DataFrame):
54
  X_scaled = scaler.fit_transform(X)
55
  return X_scaled, y, idx_list, df, scaler
56
 
57
- @fastapi_app.get("/api/train")
58
- def train_and_backtest():
59
  try:
60
  if os.path.exists(MODEL_PATH):
61
  return {"message": "Model sudah ada. Hapus rf_model.pkl & scaler.pkl untuk retrain."}
@@ -101,34 +97,30 @@ def train_and_backtest():
101
  "message": "Model trained & backtested successfully"
102
  }
103
  except Exception as e:
104
- raise HTTPException(500, f"Training failed: {str(e)}")
105
-
106
- @fastapi_app.get("/api/predict")
107
- def predict_next_move():
108
- if not os.path.exists(MODEL_PATH):
109
- raise HTTPException(400, "Model belum dilatih. Panggil /api/train terlebih dahulu.")
110
-
111
- df = fetch_data(SYMBOL, TIMEFRAME, limit=WINDOW + 10)
112
- _, _, _, df_full, scaler = prepare_dataset(df)
113
- latest_window = df_full.iloc[-WINDOW:].drop(columns=["timestamp", "next_close", "label"]).values.flatten().reshape(1, -1)
114
- latest_scaled = scaler.transform(latest_window)
115
-
116
- model = joblib.load(MODEL_PATH)
117
- pred = int(model.predict(latest_scaled)[0])
118
-
119
- return {
120
- "symbol": SYMBOL, "timeframe": TIMEFRAME,
121
- "prediction": "BUY (Hold 4h)" if pred == 1 else "HOLD/SELL",
122
- "logic": "Prediksi >0.15% movement dalam 4 jam ke depan"
123
- }
124
-
125
- # πŸ”Ή 2. Gradio UI
126
- def run_train():
127
- return train_and_backtest()
128
 
129
  def run_predict():
130
- return predict_next_move()
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
131
 
 
132
  with gr.Blocks(title="Crypto ML Predictor") as demo:
133
  gr.Markdown("## πŸ“ˆ Crypto ML Strategy (ETH/USDT 4h)")
134
  gr.Markdown("Berdasarkan paper 2105.06827v3. Membutuhkan koneksi internet untuk fetch data Binance.")
@@ -143,9 +135,5 @@ with gr.Blocks(title="Crypto ML Predictor") as demo:
143
  pred_btn.click(run_predict, outputs=pred_out)
144
  gr.Markdown("⚠️ *File model bersifat temporary. Retrain otomatis jika Space restart.*")
145
 
146
- # πŸ”Ή 3. Mount & Run Server
147
- app = gr.mount_gradio_app(fastapi_app, demo, path="/")
148
-
149
- # βœ… CRITICAL: HF Spaces butuh server yang blocking di port 7860
150
- if __name__ == "__main__":
151
- uvicorn.run(app, host="0.0.0.0", port=7860)
 
1
  import gradio as gr
 
2
  import ccxt
3
  import pandas as pd
4
  import pandas_ta as ta
 
8
  import joblib
9
  import os
10
  import warnings
 
11
 
12
  warnings.filterwarnings('ignore')
13
 
14
+ # πŸ”§ Konfigurasi
 
 
15
  SYMBOL = "ETH/USDT"
16
  TIMEFRAME = "4h"
17
  FEE = 0.0015
 
33
  df.ta.bbands(append=True)
34
  df.ta.cci(append=True)
35
  df.dropna(inplace=True)
36
+
37
  df["next_close"] = df["close"].shift(-1)
38
  df["label"] = (df["next_close"] > df["close"] * (1 + FEE)).astype(int)
39
  df.dropna(inplace=True)
 
51
  X_scaled = scaler.fit_transform(X)
52
  return X_scaled, y, idx_list, df, scaler
53
 
54
+ def run_train():
 
55
  try:
56
  if os.path.exists(MODEL_PATH):
57
  return {"message": "Model sudah ada. Hapus rf_model.pkl & scaler.pkl untuk retrain."}
 
97
  "message": "Model trained & backtested successfully"
98
  }
99
  except Exception as e:
100
+ return {"error": f"Training failed: {str(e)}"}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
101
 
102
  def run_predict():
103
+ try:
104
+ if not os.path.exists(MODEL_PATH):
105
+ return {"error": "Model belum dilatih. Klik 'Train Model & Backtest' terlebih dahulu."}
106
+
107
+ df = fetch_data(SYMBOL, TIMEFRAME, limit=WINDOW + 10)
108
+ _, _, _, df_full, scaler = prepare_dataset(df)
109
+ latest_window = df_full.iloc[-WINDOW:].drop(columns=["timestamp", "next_close", "label"]).values.flatten().reshape(1, -1)
110
+ latest_scaled = scaler.transform(latest_window)
111
+
112
+ model = joblib.load(MODEL_PATH)
113
+ pred = int(model.predict(latest_scaled)[0])
114
+
115
+ return {
116
+ "symbol": SYMBOL, "timeframe": TIMEFRAME,
117
+ "prediction": "BUY (Hold 4h)" if pred == 1 else "HOLD/SELL",
118
+ "logic": "Prediksi >0.15% movement dalam 4 jam ke depan"
119
+ }
120
+ except Exception as e:
121
+ return {"error": f"Prediction failed: {str(e)}"}
122
 
123
+ # 🎨 Gradio UI
124
  with gr.Blocks(title="Crypto ML Predictor") as demo:
125
  gr.Markdown("## πŸ“ˆ Crypto ML Strategy (ETH/USDT 4h)")
126
  gr.Markdown("Berdasarkan paper 2105.06827v3. Membutuhkan koneksi internet untuk fetch data Binance.")
 
135
  pred_btn.click(run_predict, outputs=pred_out)
136
  gr.Markdown("⚠️ *File model bersifat temporary. Retrain otomatis jika Space restart.*")
137
 
138
+ # ⏳ Queue wajib agar training tidak timeout di browser
139
+ demo.queue()