radwaamr1 commited on
Commit
7c7e513
·
verified ·
1 Parent(s): f0af22d

Upload 3 files

Browse files
Files changed (3) hide show
  1. Dockerfile +2 -0
  2. app.py +5 -47
  3. custom.py +19 -0
Dockerfile CHANGED
@@ -10,6 +10,8 @@ COPY app.py .
10
  COPY xgb_desertification_pipeline.pkl .
11
  COPY feature_names.pkl .
12
  COPY label_encoder.pkl .
 
 
13
 
14
  # Install dependencies
15
  RUN pip install --no-cache-dir -r requirements.txt
 
10
  COPY xgb_desertification_pipeline.pkl .
11
  COPY feature_names.pkl .
12
  COPY label_encoder.pkl .
13
+ COPY custom.py ./
14
+
15
 
16
  # Install dependencies
17
  RUN pip install --no-cache-dir -r requirements.txt
app.py CHANGED
@@ -3,77 +3,37 @@ from pydantic import BaseModel
3
  from typing import Dict, Any
4
  import pandas as pd
5
  import joblib
6
- import numpy as np
7
  from datetime import datetime
8
 
9
- # =========================
10
- # Custom transformer used in pipeline
11
- # =========================
12
- from sklearn.base import BaseEstimator, TransformerMixin
13
 
14
- class FeatureFilter(BaseEstimator, TransformerMixin):
15
- def __init__(self, expected_features):
16
- self.expected_features = expected_features
17
-
18
- def fit(self, X, y=None):
19
- return self
20
-
21
- def transform(self, X):
22
- X = X.copy()
23
- # keep only expected features
24
- X = X[[c for c in self.expected_features if c in X.columns]]
25
- # add missing features as NaN
26
- for c in self.expected_features:
27
- if c not in X.columns:
28
- X[c] = np.nan
29
- return X[self.expected_features]
30
-
31
- # =========================
32
- # Load model artifacts
33
- # =========================
34
  xgb_pipe = joblib.load("xgb_desertification_pipeline.pkl")
35
  label_encoder = joblib.load("label_encoder.pkl")
36
  expected_features = joblib.load("feature_names.pkl")
37
 
38
- # =========================
39
  # FastAPI app
40
- # =========================
41
  app = FastAPI(title="Desertification Prediction API")
42
 
43
- # =========================
44
- # Input schema
45
- # =========================
46
  class PredictionRequest(BaseModel):
47
  success: bool
48
  features: Dict[str, Any]
49
  metadata: Dict[str, Any] | None = None
50
 
51
- # =========================
52
- # Health check endpoint
53
- # =========================
54
  @app.get("/health")
55
  def health():
56
  return {"status": "ok", "timestamp": datetime.now().isoformat()}
57
 
58
- # =========================
59
- # Prediction endpoint
60
- # =========================
61
  @app.post("/predict")
62
  def predict(request: PredictionRequest):
63
-
64
- # Convert features to DataFrame
65
  input_df = pd.DataFrame([request.features])
66
-
67
- # Ensure correct feature order & missing features handling
68
  input_df = input_df.reindex(columns=expected_features)
69
-
70
- # Predict
71
  pred_encoded = xgb_pipe.predict(input_df)[0]
72
  pred_label = label_encoder.inverse_transform([pred_encoded])[0]
73
  pred_proba = float(xgb_pipe.predict_proba(input_df).max())
74
-
75
- # Build response
76
- response = {
77
  "success": True,
78
  "prediction": {
79
  "desertification_level": pred_label,
@@ -81,5 +41,3 @@ def predict(request: PredictionRequest):
81
  },
82
  "metadata": request.metadata
83
  }
84
-
85
- return response
 
3
  from typing import Dict, Any
4
  import pandas as pd
5
  import joblib
 
6
  from datetime import datetime
7
 
8
+ # Import custom class BEFORE loading pipeline
9
+ from custom import FeatureFilter
 
 
10
 
11
+ # Load artifacts
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
12
  xgb_pipe = joblib.load("xgb_desertification_pipeline.pkl")
13
  label_encoder = joblib.load("label_encoder.pkl")
14
  expected_features = joblib.load("feature_names.pkl")
15
 
 
16
  # FastAPI app
 
17
  app = FastAPI(title="Desertification Prediction API")
18
 
19
+ # Schema
 
 
20
  class PredictionRequest(BaseModel):
21
  success: bool
22
  features: Dict[str, Any]
23
  metadata: Dict[str, Any] | None = None
24
 
 
 
 
25
  @app.get("/health")
26
  def health():
27
  return {"status": "ok", "timestamp": datetime.now().isoformat()}
28
 
 
 
 
29
  @app.post("/predict")
30
  def predict(request: PredictionRequest):
 
 
31
  input_df = pd.DataFrame([request.features])
 
 
32
  input_df = input_df.reindex(columns=expected_features)
 
 
33
  pred_encoded = xgb_pipe.predict(input_df)[0]
34
  pred_label = label_encoder.inverse_transform([pred_encoded])[0]
35
  pred_proba = float(xgb_pipe.predict_proba(input_df).max())
36
+ return {
 
 
37
  "success": True,
38
  "prediction": {
39
  "desertification_level": pred_label,
 
41
  },
42
  "metadata": request.metadata
43
  }
 
 
custom.py ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import numpy as np
2
+ from sklearn.base import BaseEstimator, TransformerMixin
3
+
4
+ class FeatureFilter(BaseEstimator, TransformerMixin):
5
+ def __init__(self, expected_features):
6
+ self.expected_features = expected_features
7
+
8
+ def fit(self, X, y=None):
9
+ return self
10
+
11
+ def transform(self, X):
12
+ X = X.copy()
13
+ # keep only expected features
14
+ X = X[[c for c in self.expected_features if c in X.columns]]
15
+ # add missing features as NaN
16
+ for c in self.expected_features:
17
+ if c not in X.columns:
18
+ X[c] = np.nan
19
+ return X[self.expected_features]