Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+samples/00001.wav filter=lfs diff=lfs merge=lfs -text
+samples/00002.wav filter=lfs diff=lfs merge=lfs -text

app.py

@@ -2,40 +2,64 @@ import torch
 import gradio as gr
 from model import ECAPA_gender
 
+
+SAMPLE_AUDIO = [
+    ("Sample 1", "samples/00001.wav"),
+    ("Sample 2", "samples/00002.wav"),
+]
+
 model = ECAPA_gender.from_pretrained("Beijuka/voice-gender-classifier")
 model.eval()
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 model.to(device)
 
+
 def predict_gender_confidence(audio_file):
     if audio_file is None:
         return "No audio provided"
-    
+
     try:
-        # Load audio
-        audio = model.load_audio(audio_file.name if hasattr(audio_file, "name") else audio_file)
+        path = audio_file if isinstance(audio_file, str) else getattr(audio_file, "name", None)
+        if not path:
+            return "No audio path provided"
+
+        audio = model.load_audio(path)
         audio = audio.to(device)
-        
-        # Forward pass
+
         with torch.no_grad():
             logits = model.forward(audio)
-            probs = torch.softmax(logits, dim=1).cpu().numpy()[0] 
+            probs = torch.softmax(logits, dim=1).cpu().numpy()[0]
             pred_idx = logits.argmax(dim=1).item()
             gender_pred = model.pred2gender[pred_idx].capitalize()
-            confidence = probs[pred_idx] * 100 
-        
+            confidence = probs[pred_idx] * 100
+
         return f"{gender_pred} — {confidence:.1f}% confidence"
-    
+
     except Exception as e:
         return f"Error: {e}"
 
-iface = gr.Interface(
-    fn=predict_gender_confidence,
-    inputs=gr.Audio(type="filepath", label="Upload audio file", sources=["upload"]),
-    outputs=gr.Textbox(label="Predicted Gender with Confidence"),
-    title="Voice Gender Classifier",
-    description="Upload an audio file and the model predicts speaker gender with confidence.",
-    allow_flagging="never"
-)
 
-iface.launch(share=True)
+with gr.Blocks(title="Voice Gender Classifier") as demo:
+    gr.Markdown("""
+    ## Voice Gender Classifier
+    Upload or record a short audio clip to predict speaker gender. Try the built-in samples if you need test audio.
+    """)
+
+    audio_input = gr.Audio(
+        sources=["upload", "microphone"],
+        type="filepath",
+        label="Upload or record audio",
+    )
+    prediction = gr.Textbox(label="Prediction", interactive=False)
+
+    gr.Examples(
+        examples=[path for _, path in SAMPLE_AUDIO],
+        inputs=audio_input,
+        outputs=prediction,
+        fn=predict_gender_confidence,
+        label="Try sample audios",
+    )
+
+    audio_input.change(fn=predict_gender_confidence, inputs=audio_input, outputs=prediction)
+
+demo.launch(share=True)

requirements.txt

@@ -1,7 +1,8 @@
 torch
 torchaudio
 pysoundfile
-gradio==4.44.1
-huggingface_hub==0.24.5
+huggingface_hub>=0.23.0
 safetensors
-fastapi<0.113.0
+gradio==4.31.1
+gradio-client==0.16.3
+fastapi>=0.110,<0.116

samples/00001.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c71446ec2322ae5aa480a5e8865dfdc3535b4bac238dd87e3cb7edd06dab89fd
+size 267566

samples/00002.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ecdf81edc906e8941c6dfbad12b8215b2adb9f1ace079e97014073687764e63c
+size 238126