Initial version

app.py

@@ -0,0 +1,40 @@
+import gradio as gr
+from transformers import pipeline
+
+model_name = "juliensimon/wav2vec2-conformer-rel-pos-large-finetuned-speech-commands"
+
+p = pipeline("audio-classification", model=model_name)
+
+def process(file):
+        pred = p(file)
+        labels = dict()
+        for l in pred:
+            labels[l['label']]=l['score']
+        return labels
+
+# Gradio inputs
+mic = gr.inputs.Audio(source='microphone', type='filepath', label='Speech input', optional=True)
+
+# Gradio outputs
+keyword = gr.outputs.Label(num_top_classes=3)
+
+description = "This Space showcases a wav2vec2-conformer-rel-pos-large model fine-tuned for audio classification on the speech_commands dataset. \n \n It can spot one of the following keywords: 'Yes', 'No', 'Up', 'Down', 'Left', 'Right', 'On', 'Off', 'Stop', 'Go', 'Zero', 'One', 'Two', 'Three', 'Four', 'Five', 'Six', 'Seven', 'Eight', 'Nine', 'Bed', 'Bird', 'Cat', 'Dog', 'Happy', 'House', 'Marvin', 'Sheila', 'Tree', 'Wow', 'Backward', 'Forward', 'Follow', 'Learn', 'Visual'."
+
+iface = gr.Interface(
+	theme='huggingface',
+	description=description,
+	fn=process,
+        layout='horizontal',
+	inputs=[mic],
+	outputs=[keyword],
+	examples=[
+		['backward16k.wav'],
+		['happy16k.wav'],
+		['marvin16k.wav'],
+		['seven16k.wav'],
+		['stop16k.wav'],
+		['up16k.wav'],
+	],
+    allow_flagging=False
+)
+iface.launch()

requirements.txt

@@ -0,0 +1,3 @@
+torch
+transformers
+librosa