init

app.py

@@ -1,7 +1,83 @@
 import gradio as gr
+import soundfile as sf
+import torch
+from transformers import Qwen2_5OmniForConditionalGeneration, Qwen2_5OmniProcessor
 
-def greet(name):
-    return "Hello " + name + "!!"
 
-demo = gr.Interface(fn=greet, inputs="text", outputs="text")
-demo.launch()
+MODEL_ID = "Qwen/Qwen2.5-Omni-7B" if False else "Qwen/Qwen2.5-Omni-7B"  # keep explicit string
+
+model = Qwen2_5OmniForConditionalGeneration.from_pretrained(
+    MODEL_ID,
+    torch_dtype="auto",
+    device_map="auto",
+)
+model.disable_talker()
+
+processor = Qwen2_5OmniProcessor.from_pretrained(MODEL_ID, trust_remote_code=True)
+
+
+def run_omni(audio_path: str, instruction: str, max_tokens: int = 512) -> str:
+    if not audio_path:
+        return "Please upload an audio file."
+
+    system_text = (
+        "You are Qwen, a virtual human developed by the Qwen Team, Alibaba Group, "
+        "capable of perceiving auditory and visual inputs, as well as generating text and speech."
+    )
+    conversation = [
+        {
+            "role": "system",
+            "content": [{"type": "text", "text": system_text}],
+        },
+        {
+            "role": "user",
+            "content": [
+                {"type": "audio", "audio_url": audio_path},
+                {"type": "text", "text": instruction},
+            ],
+        },
+    ]
+
+    text = processor.apply_chat_template(conversation, add_generation_prompt=True, tokenize=False)
+
+    audio, sr = sf.read(audio_path)
+    audios = [(audio, sr)]
+
+    inputs = processor(
+        text=text,
+        audio=audios,
+        images=[],
+        videos=[],
+        return_tensors="pt",
+        padding=True,
+    )
+    inputs = inputs.to(model.device)
+
+    output_ids = model.generate(**inputs, max_new_tokens=int(max_tokens))
+    response = processor.batch_decode(
+        output_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False
+    )[0]
+    return response
+
+
+with gr.Blocks(title="Qwen2.5 Omni (Audio) Demo") as demo:
+    gr.Markdown("# Qwen2.5-Omni (Audio) Demo")
+    gr.Markdown("Upload an audio file and provide an instruction for the model.")
+
+    with gr.Row():
+        with gr.Column():
+            audio_input = gr.Audio(type="filepath", label="Upload Audio")
+            instruction = gr.Textbox(
+                label="Instruction",
+                value="Transcribe the audio, then summarize it in one sentence.",
+            )
+            max_tokens = gr.Slider(128, 2048, value=512, step=64, label="Max Output Tokens")
+            submit_btn = gr.Button("Run", variant="primary")
+        with gr.Column():
+            output_text = gr.Textbox(label="Response", lines=14)
+
+    submit_btn.click(run_omni, [audio_input, instruction, max_tokens], output_text)
+
+
+if __name__ == "__main__":
+    demo.queue().launch(share=False, ssr_mode=False)

requirements.txt

@@ -0,0 +1,6 @@
+gradio>=4.0.0
+torch>=2.1.0
+transformers>=4.43.0
+accelerate>=0.30.0
+soundfile>=0.12.1
+