Create app.py

app.py

@@ -0,0 +1,44 @@
+import gradio as gr
+import os
+import shutil
+import spaces
+
+# we will clone the repo and install the dependencies
+os.system('git lfs install')
+os.system('git clone https://huggingface.co/jadechoghari/qa-mdt')
+os.system('pip install -r qa_mdt/requirements.txt')
+os.system('pip install xformers==0.0.26.post1')
+os.system('pip install torchlibrosa==0.0.9 librosa==0.9.2')
+os.system('pip install -q pytorch_lightning==2.1.3 torchlibrosa==0.0.9 librosa==0.9.2 ftfy==6.1.1 braceexpand')
+os.system('pip install torch==2.3.0+cu121 torchvision==0.18.0+cu121 torchaudio==2.3.0 --index-url https://download.pytorch.org/whl/cu121')
+
+# only then import the necessary modules from qa_mdt
+from qa_mdt.pipeline import MOSDiffusionPipeline
+
+
+pipe = MOSDiffusionPipeline()
+
+# this runs the pipeline with user input and saves the output as 'awesome.wav'
+@spaces.GPU()
+def generate_waveform(description):
+    pipe(description)
+
+    generated_file_path = "./awesome.wav"
+
+    if os.path.exists(generated_file_path):
+        return generated_file_path
+    else:
+        return "Error: Failed to generate the waveform."
+
+# gradio interface
+iface = gr.Interface(
+    fn=generate_waveform,
+    inputs=gr.inputs.Textbox(lines=2, placeholder="Enter a music description here..."),  # Text input for description
+    outputs=gr.outputs.File(label="Download Generated WAV file"),  # File output for download
+    title="Flux Music Diffusion Pipeline",
+    description="Enter a music description, and the model will generate a corresponding audio waveform. Download the output as 'awesome.wav'."
+)
+
+# Launch the Gradio app
+if __name__ == "__main__":
+    iface.launch()