Update app.py

app.py

@@ -2,6 +2,9 @@ import gradio as gr
 import numpy as np
 import tensorflow as tf
 import joblib
+import requests
+import json
+import os
 
 # Load the model and scalers
 lstm_model = tf.keras.models.load_model('retrained_lstm_model.h5', compile=False)
@@ -12,21 +15,29 @@ scaler_y = joblib.load('scaler_y.joblib')
 # Define window size
 window_size = 7
 
+# Function to query Gemma model using Hugging Face Inference API
+def query_gemma(prompt):
+    API_URL = "https://api-inference.huggingface.co/models/google/gemma-2-2b-it"
+    headers = {"Authorization": f"Bearer {os.environ.get('HF_TOKEN', '')}"}
+    
+    payload = {
+        "inputs": prompt,
+        "parameters": {
+            "max_new_tokens": 300,
+            "temperature": 0.7,
+            "top_p": 0.9,
+            "do_sample": True
+        }
+    }
+    
+    try:
+        response = requests.post(API_URL, headers=headers, json=payload)
+        return response.json()[0]["generated_text"].replace(prompt, "").strip()
+    except Exception as e:
+        return f"Error generating interpretation: {str(e)}"
+
 # Create Gradio app
 with gr.Blocks(fill_height=True, title="Rainfall Prediction with LSTM and Gemma 2") as demo:
-    # Set up Gemma login
-    with gr.Sidebar():
-        gr.Markdown("# Inference Provider")
-        gr.Markdown("This Space showcases the google/gemma-2-2b-it model, served by the Nebius API. Sign in with your Hugging Face account to use this API.")
-        login_button = gr.LoginButton("Sign in with Hugging Face")
-        model_provider = gr.Dropdown(
-            ["nebius"], 
-            label="Provider", 
-            value="nebius", 
-            visible=False
-        )
-
-    # Main content
     gr.Markdown("# Rainfall Prediction with LSTM and Gemma 2 Interpretation")
     gr.Markdown("Enter temperature, humidity, and sunshine duration to predict rainfall.")
     
@@ -41,13 +52,6 @@ with gr.Blocks(fill_height=True, title="Rainfall Prediction with LSTM and Gemma
             rainfall_output = gr.Textbox(label="Predicted Rainfall (mm)")
             interpretation_output = gr.Markdown(label="Interpretation by Gemma 2")
 
-    # Load Gemma model using gr.load - this returns a function to call the model
-    gemma_model = gr.load(
-        "models/google/gemma-2-2b-it",
-        provider=model_provider,
-        accept_token=login_button
-    )
-
     def predict_rainfall(tavg, rh_avg, ss):
         # Create sequential input
         user_input = np.array([[tavg, rh_avg, ss]] * window_size)
@@ -77,18 +81,8 @@ with gr.Blocks(fill_height=True, title="Rainfall Prediction with LSTM and Gemma
         Provide a concise interpretation with practical implications.
         """
 
-        # Request Gemma model for interpretation
-        try:
-            # Call the loaded model directly with the prompt
-            response = gemma_model(
-                prompt,
-                max_new_tokens=300,
-                temperature=0.7,
-                top_p=0.9
-            )
-            interpretation = response
-        except Exception as e:
-            interpretation = f"Could not generate interpretation. Please make sure you're signed in with your Hugging Face account. Error: {str(e)}"
+        # Get interpretation from Gemma
+        interpretation = query_gemma(prompt)
         
         return f"{rainfall_mm:.2f}", interpretation