Update app.py

app.py

@@ -1,64 +1,68 @@
 import gradio as gr
 import torch
-import torch.nn as nn
-from transformers import AutoTokenizer, AutoModel
+from modeling import BERTMultiLabel
+from transformers import AutoTokenizer
 
-LABELS = ["anger", "fear", "joy", "sadness", "surprise"]   
 
+LABELS = ["anger", "fear", "joy", "sadness", "surprise"]
 
-class BERTMultiLabel(nn.Module):
-    def __init__(self, model_name="microsoft/deberta-v3-base", num_labels=5):
-        super().__init__()
-        self.bert = AutoModel.from_pretrained(model_name)
-        hidden = self.bert.config.hidden_size
-        self.dropout = nn.Dropout(0.2)
-        self.classifier = nn.Linear(hidden, num_labels)
+# ---- Load tokenizer ----
+tokenizer = AutoTokenizer.from_pretrained("./")
 
-    def forward(self, input_ids, attention_mask):
-        outputs = self.bert(input_ids=input_ids, attention_mask=attention_mask)
-        cls = outputs.last_hidden_state[:, 0]
-        cls = self.dropout(cls)
-        return self.classifier(cls)
 
-MODEL_PATH = "./"  
-
-tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)
+model = BERTMultiLabel(model_name="microsoft/deberta-v3-base", num_labels=len(LABELS))
+state = torch.load("pytorch_model.bin", map_location="cpu")
+model.load_state_dict(state)
+model.eval()
 
-model = BERTMultiLabel(
-    model_name="microsoft/deberta-v3-base",
-    num_labels=len(LABELS)
-)
+-
+def predict(text):
+    if text.strip() == "":
+        return {"error": "Please enter text"}
 
-state_dict = torch.load("pytorch_model.bin", map_location="cpu")
-model.load_state_dict(state_dict)
-model.eval()
+    enc = tokenizer(
+        text,
+        truncation=True,
+        padding="max_length",
+        max_length=128,
+        return_tensors="pt"
+    )
 
-def predict_mood(text):
-    enc = tokenizer(text, truncation=True, padding="max_length", max_length=128, return_tensors="pt")
     with torch.no_grad():
         logits = model(enc["input_ids"], enc["attention_mask"])
         probs = torch.sigmoid(logits).flatten().tolist()
 
     result = {label: float(f"{p:.4f}") for label, p in zip(LABELS, probs)}
-    return result
+    top_label = LABELS[int(torch.argmax(torch.tensor(probs)))]
+
+    return {
+        "Predicted Mood": top_label,
+        "Scores": result
+    }
+
+
+with gr.Blocks(title="Mood Detector") as app:
+    gr.Markdown("# 🧠 DeBERTa Mood Detector")
+    gr.Markdown("Enter any text to analyze the emotional tone.")
 
-with gr.Blocks() as app:
-    gr.Markdown("# 🧠 Mood Detector – DeBERTa-V3")
-    
     with gr.Row():
+        # LEFT PANEL
         with gr.Column(scale=1):
             gr.Markdown("""
-            ### 📌 Model Details
-            - **Architecture:** DeBERTa-V3 Base  
-            - **Task:** Multi-label emotional mood detection  
-            - **Labels:** anger, fear, joy, sadness, surprise  
-            - **Framework:** PyTorch + Transformers  
+            ## 📌 Model Info
+            - **Architecture:** DeBERTa-v3 Base  
+            - **Task:** Multi-label emotion classification  
+            - **Trained by:** You  
+            - **Dataset:** Custom text emotion dataset  
+            - **Output:** anger, fear, joy, sadness, surprise  
             """)
         
+        # RIGHT PANEL
         with gr.Column(scale=2):
-            text = gr.Textbox(label="Enter your text")
-            output = gr.Label(label="Predicted Mood Probabilities")
-            btn = gr.Button("Analyze Mood")
-            btn.click(fn=predict_mood, inputs=text, outputs=output)
+            user_input = gr.Textbox(label="Enter text here", placeholder="Type something like: 'I feel so happy today!'")
+            analyze_button = gr.Button("Analyze Mood")
+            output = gr.JSON(label="Result")
+
+            analyze_button.click(predict, inputs=user_input, outputs=output)
 
 app.launch()