Update: Add custom NPKPredictionModel implementation

README.md

@@ -0,0 +1,33 @@
+---
+tags:
+- npk-prediction
+- xgboost
+---
+
+# NPK Prediction Model
+
+This model predicts the Nitrogen, Phosphorus, and Potassium needs for crops based on various input features.
+
+## Usage
+
+```python
+from transformers import AutoConfig, AutoModel
+
+# Load the model
+config = AutoConfig.from_pretrained("GodfreyOwino/NPK_prediction_model1")
+model = AutoModel.from_config(config)
+
+input_data = {
+    'crop_name': ['maize (corn)'],
+    'target_yield': [150],
+    'field_size': [10],
+    'ph': [6.5],
+    'organic_carbon': [1.2],
+    'nitrogen': [0.15],
+    'phosphorus': [20],
+    'potassium': [150],
+    'soil_moisture': [30]
+}
+
+prediction = model(input_data)
+print(prediction)

config.json

@@ -0,0 +1,8 @@
+{
+  "model_type": "npk",
+  "architectures": ["NPKPredictionModel"],
+  "auto_map": {
+  "AutoConfig": "modeling_npk.NPKConfig",
+  "AutoModel": "modeling_npk.NPKPredictionModel"
+  }
+  }

label_encoder.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac25981a59fd92e29772ecf03261fa9992cc2df471d6fd05d77977930203066c
+size 1233

modeling_npk.py

@@ -0,0 +1,62 @@
+
+import pickle
+import pandas as pd
+from transformers import PreTrainedModel, PretrainedConfig
+from huggingface_hub import hf_hub_download
+
+class NPKConfig(PretrainedConfig):
+    model_type = "npk"
+    
+    def __init__(self, **kwargs):
+        super().__init__(**kwargs)
+
+class NPKPredictionModel(PreTrainedModel):
+    config_class = NPKConfig
+    
+    def __init__(self, config):
+        super().__init__(config)
+        self.xgb_model = None
+        self.label_encoder = None
+    
+    def forward(self, inputs):
+        # Preprocess inputs
+        processed_inputs = {}
+        for key, value in inputs.items():
+            if isinstance(value, list):
+                processed_inputs[key] = value[0] if value else None
+            else:
+                processed_inputs[key] = value
+
+        crop_name = processed_inputs['crop_name']
+        processed_inputs['crop_name'] = self.label_encoder.transform([crop_name])[0]
+        
+        input_df = pd.DataFrame([processed_inputs])
+        
+        # Make prediction
+        prediction = self.xgb_model.predict(input_df)
+        
+        return {
+            'Nitrogen Need': float(prediction[0][0]),
+            'Phosphorus Need': float(prediction[0][1]),
+            'Potassium Need': float(prediction[0][2])
+        }
+    
+    @classmethod
+    def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
+        config = kwargs.pop("config", None)
+        if config is None:
+            config = NPKConfig.from_pretrained(pretrained_model_name_or_path, **kwargs)
+        
+        model = cls(config)
+        
+        # Load the XGBoost model and label encoder
+        xgb_path = hf_hub_download(repo_id=pretrained_model_name_or_path, filename="npk_prediction_model.pkl")
+        le_path = hf_hub_download(repo_id=pretrained_model_name_or_path, filename="label_encoder.pkl")
+        
+        with open(xgb_path, 'rb') as f:
+            model.xgb_model = pickle.load(f)
+        
+        with open(le_path, 'rb') as f:
+            model.label_encoder = pickle.load(f)
+        
+        return model

npk_prediction_model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03507cafdc8061c0a14f23d1791e4d7c908752e68c5b1f330f0c7854a398a640
+size 129759981