Update app.py
Browse files
app.py
CHANGED
|
@@ -1,34 +1,27 @@
|
|
| 1 |
from transformers import AutoProcessor, MusicgenForConditionalGeneration
|
| 2 |
-
from IPython.display import Audio
|
| 3 |
-
import scipy
|
| 4 |
import torch
|
| 5 |
import streamlit as st
|
| 6 |
|
| 7 |
-
|
| 8 |
def mu_gen(prompt):
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
device = torch.device("cpu")
|
| 13 |
-
model.to(device)
|
| 14 |
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
padding=True,
|
| 18 |
-
return_tensors="pt",
|
| 19 |
-
)
|
| 20 |
-
|
| 21 |
-
inputs = {key: value.to(device) for key, value in inputs.items()}
|
| 22 |
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
|
|
|
|
|
|
| 26 |
|
| 27 |
-
|
| 28 |
-
result = Audio(audio_values[0].numpy(), rate=sampling_rate)
|
| 29 |
|
| 30 |
-
|
|
|
|
|
|
|
| 31 |
|
|
|
|
| 32 |
|
| 33 |
def main():
|
| 34 |
st.title("Text to Music Generator")
|
|
@@ -39,12 +32,12 @@ def main():
|
|
| 39 |
if st.button("Generate Music"):
|
| 40 |
if prompt:
|
| 41 |
# Call the mu_gen function to generate music
|
| 42 |
-
generated_music = mu_gen(prompt)
|
| 43 |
|
| 44 |
# Display the generated audio
|
| 45 |
-
st.audio(generated_music, format="audio/wav")
|
| 46 |
else:
|
| 47 |
st.warning("Please enter a text prompt.")
|
| 48 |
|
| 49 |
if __name__ == "__main__":
|
| 50 |
-
main()
|
|
|
|
| 1 |
from transformers import AutoProcessor, MusicgenForConditionalGeneration
|
|
|
|
|
|
|
| 2 |
import torch
|
| 3 |
import streamlit as st
|
| 4 |
|
|
|
|
| 5 |
def mu_gen(prompt):
|
| 6 |
+
processor = AutoProcessor.from_pretrained("facebook/musicgen-small")
|
| 7 |
+
model = MusicgenForConditionalGeneration.from_pretrained("facebook/musicgen-small")
|
|
|
|
|
|
|
|
|
|
| 8 |
|
| 9 |
+
device = torch.device("cpu")
|
| 10 |
+
model.to(device)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
|
| 12 |
+
inputs = processor(
|
| 13 |
+
text=[str(prompt)],
|
| 14 |
+
padding=True,
|
| 15 |
+
return_tensors="pt",
|
| 16 |
+
)
|
| 17 |
|
| 18 |
+
inputs = {key: value.to(device) for key, value in inputs.items()}
|
|
|
|
| 19 |
|
| 20 |
+
# Generate audio on CPU
|
| 21 |
+
audio_values = model.generate(**inputs, max_new_tokens=256)
|
| 22 |
+
sampling_rate = model.config.audio_encoder.sampling_rate
|
| 23 |
|
| 24 |
+
return audio_values, sampling_rate
|
| 25 |
|
| 26 |
def main():
|
| 27 |
st.title("Text to Music Generator")
|
|
|
|
| 32 |
if st.button("Generate Music"):
|
| 33 |
if prompt:
|
| 34 |
# Call the mu_gen function to generate music
|
| 35 |
+
generated_music, sampling_rate = mu_gen(prompt)
|
| 36 |
|
| 37 |
# Display the generated audio
|
| 38 |
+
st.audio(generated_music[0].numpy(), format="audio/wav", sample_rate=sampling_rate)
|
| 39 |
else:
|
| 40 |
st.warning("Please enter a text prompt.")
|
| 41 |
|
| 42 |
if __name__ == "__main__":
|
| 43 |
+
main()
|