First version

app.py

@@ -1,8 +1,17 @@
 import gradio as gr
+import torch
 from model import ECAPA_gender
 
-def greet(audio):
-    print(audio)
+model = ECAPA_gender({"C": 1024})
+model.load_state_dict(torch.load("gender_classifier.model", map_location="cpu"))
 
-demo = gr.Interface(fn=greet, inputs="audio", outputs="text")
+model.eval()
+
+def predict_gender(filepath):
+    with torch.no_grad():
+        output = model.predict(filepath)
+    return output
+
+audio_component = gr.Audio(type='filepath', label="Upload your audio file here")
+demo = gr.Interface(fn=predict_gender, inputs=audio_component, outputs="text")
 demo.launch()

gender_classifier.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0e87253fe40aeef61b33e0fbe0859d9407e8bed92c8512cbaa28593b5f98e13
+size 62042234

model.py

@@ -89,7 +89,7 @@ class PreEmphasis(torch.nn.Module):
         return F.conv1d(input, self.flipped_filter).squeeze(1)
 
 
-class ECAPA_gender(nn.Module, PyTorchModelHubMixin):
+class ECAPA_gender(nn.Module):
     def __init__(self, config):
         super(ECAPA_gender, self).__init__()
         self.config = config