Update app.py

app.py

@@ -1,26 +1,11 @@
 import gradio as gr
 import torch
-import numpy as np
 import librosa
-
 from transformers import AutoFeatureExtractor
-
 from model import MMSForMultilingualSER
 
 MODEL_ID = "E-motionAssistant/mms-300m-multilingual-ser"
 
-# Load feature extractor + model
-feature_extractor = AutoFeatureExtractor.from_pretrained(MODEL_ID)
-
-emotion_model = MMSForMultilingualSER.from_pretrained(
-    MODEL_ID,
-    ignore_mismatched_sizes=True
-)
-
-emotion_model.eval()
-
-
-# Emotion labels (adjust if different)
 emotion_labels = [
     "neutral",
     "happy",
@@ -29,10 +14,24 @@ emotion_labels = [
     "fear"
 ]
 
+device = "cpu"
 
-def detect_emotion(audio_file):
+print("Loading model...")
 
-    speech, sr = librosa.load(audio_file, sr=16000)
+feature_extractor = AutoFeatureExtractor.from_pretrained(MODEL_ID)
+
+emotion_model = MMSForMultilingualSER.from_pretrained(
+    MODEL_ID
+).to(device)
+
+emotion_model.eval()
+
+print("Model loaded")
+
+
+def detect_emotion(audio):
+
+    speech, sr = librosa.load(audio, sr=16000)
 
     inputs = feature_extractor(
         speech,
@@ -52,20 +51,17 @@ with gr.Blocks() as demo:
 
     gr.Markdown("# Emotion Regulation Assistant")
 
-    gr.TabbedInterface(
-        [
-            gr.Interface(
-                fn=detect_emotion,
-                inputs=gr.Audio(type="filepath"),
-                outputs=gr.Textbox(),
-                title="Emotion Detection"
-            )
-        ],
-
-        [
-            "Emotion Detection"
-        ]
-    )
+    with gr.Tab("Emotion Detection"):
+
+        audio_input = gr.Audio(type="filepath")
+        output = gr.Textbox(label="Detected Emotion")
+
+        btn = gr.Button("Detect Emotion")
 
+        btn.click(
+            fn=detect_emotion,
+            inputs=audio_input,
+            outputs=output
+        )
 
 demo.launch()