Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,4 +1,3 @@
|
|
1 |
-
|
2 |
import gradio as gr
|
3 |
import os
|
4 |
import torch
|
@@ -14,15 +13,13 @@ import tempfile
|
|
14 |
from dotenv import load_dotenv
|
15 |
import spaces
|
16 |
|
17 |
-
|
18 |
load_dotenv()
|
19 |
hf_token = os.getenv("HF_TOKEN")
|
20 |
|
21 |
-
|
22 |
# ---------------------------------------------------------------------
|
23 |
# Load Llama 3 Pipeline with Zero GPU (Encapsulated)
|
24 |
# ---------------------------------------------------------------------
|
25 |
-
@spaces.GPU(duration=300)
|
26 |
def generate_script(user_prompt: str, model_id: str, token: str):
|
27 |
try:
|
28 |
tokenizer = AutoTokenizer.from_pretrained(model_id, use_auth_token=token)
|
@@ -50,7 +47,6 @@ def generate_script(user_prompt: str, model_id: str, token: str):
|
|
50 |
# ---------------------------------------------------------------------
|
51 |
# Load MusicGen Model (Encapsulated)
|
52 |
# ---------------------------------------------------------------------
|
53 |
-
|
54 |
@spaces.GPU(duration=300)
|
55 |
def generate_audio(prompt: str, audio_length: int):
|
56 |
try:
|
@@ -60,7 +56,7 @@ def generate_audio(prompt: str, audio_length: int):
|
|
60 |
musicgen_model.to("cuda")
|
61 |
inputs = musicgen_processor(text=[prompt], padding=True, return_tensors="pt")
|
62 |
outputs = musicgen_model.generate(**inputs, max_new_tokens=audio_length)
|
63 |
-
musicgen_model.to("cpu")
|
64 |
|
65 |
sr = musicgen_model.config.audio_encoder.sampling_rate
|
66 |
audio_data = outputs[0, 0].cpu().numpy()
|
@@ -69,7 +65,7 @@ def generate_audio(prompt: str, audio_length: int):
|
|
69 |
output_path = f"{tempfile.gettempdir()}/generated_audio.wav"
|
70 |
write(output_path, sr, normalized_audio)
|
71 |
|
72 |
-
return output_path
|
73 |
except Exception as e:
|
74 |
return f"Error generating audio: {e}"
|
75 |
|
@@ -93,7 +89,8 @@ with gr.Blocks() as demo:
|
|
93 |
gr.Markdown("""
|
94 |
# 🎙️ AI-Powered Radio Imaging Studio 🚀
|
95 |
### Create stunning **radio promos** with **Llama 3** and **MusicGen**
|
96 |
-
🔥 **Zero GPU** integration for efficiency and ease!
|
|
|
97 |
""")
|
98 |
|
99 |
# Script Generation Section
|
@@ -127,20 +124,21 @@ with gr.Blocks() as demo:
|
|
127 |
maximum=1024,
|
128 |
step=64,
|
129 |
value=512,
|
130 |
-
info="Select the desired audio token length."
|
|
|
|
|
|
|
|
|
|
|
|
|
131 |
)
|
132 |
-
audio_output = gr.Audio(
|
133 |
-
label="🎶 Generated Audio File",
|
134 |
-
type="filepath", # Correct parameter to handle file paths
|
135 |
-
interactive=False
|
136 |
-
)
|
137 |
-
|
138 |
|
139 |
# Footer
|
140 |
gr.Markdown("""
|
141 |
<br><hr>
|
142 |
<p style="text-align: center; font-size: 0.9em;">
|
143 |
-
Created with ❤️ by <a href="https://bilsimaging.com" target="_blank">bilsimaging.com</a>
|
|
|
144 |
</p>
|
145 |
""", elem_id="footer")
|
146 |
|
|
|
|
|
1 |
import gradio as gr
|
2 |
import os
|
3 |
import torch
|
|
|
13 |
from dotenv import load_dotenv
|
14 |
import spaces
|
15 |
|
|
|
16 |
load_dotenv()
|
17 |
hf_token = os.getenv("HF_TOKEN")
|
18 |
|
|
|
19 |
# ---------------------------------------------------------------------
|
20 |
# Load Llama 3 Pipeline with Zero GPU (Encapsulated)
|
21 |
# ---------------------------------------------------------------------
|
22 |
+
@spaces.GPU(duration=300)
|
23 |
def generate_script(user_prompt: str, model_id: str, token: str):
|
24 |
try:
|
25 |
tokenizer = AutoTokenizer.from_pretrained(model_id, use_auth_token=token)
|
|
|
47 |
# ---------------------------------------------------------------------
|
48 |
# Load MusicGen Model (Encapsulated)
|
49 |
# ---------------------------------------------------------------------
|
|
|
50 |
@spaces.GPU(duration=300)
|
51 |
def generate_audio(prompt: str, audio_length: int):
|
52 |
try:
|
|
|
56 |
musicgen_model.to("cuda")
|
57 |
inputs = musicgen_processor(text=[prompt], padding=True, return_tensors="pt")
|
58 |
outputs = musicgen_model.generate(**inputs, max_new_tokens=audio_length)
|
59 |
+
musicgen_model.to("cpu")
|
60 |
|
61 |
sr = musicgen_model.config.audio_encoder.sampling_rate
|
62 |
audio_data = outputs[0, 0].cpu().numpy()
|
|
|
65 |
output_path = f"{tempfile.gettempdir()}/generated_audio.wav"
|
66 |
write(output_path, sr, normalized_audio)
|
67 |
|
68 |
+
return output_path
|
69 |
except Exception as e:
|
70 |
return f"Error generating audio: {e}"
|
71 |
|
|
|
89 |
gr.Markdown("""
|
90 |
# 🎙️ AI-Powered Radio Imaging Studio 🚀
|
91 |
### Create stunning **radio promos** with **Llama 3** and **MusicGen**
|
92 |
+
🔥 **Zero GPU** integration for efficiency and ease!
|
93 |
+
❤️ A huge thanks to the **Hugging Face community** for making this possible.
|
94 |
""")
|
95 |
|
96 |
# Script Generation Section
|
|
|
124 |
maximum=1024,
|
125 |
step=64,
|
126 |
value=512,
|
127 |
+
info="Select the desired audio token length."
|
128 |
+
)
|
129 |
+
generate_audio_button = gr.Button("Generate Audio 🎶")
|
130 |
+
audio_output = gr.Audio(
|
131 |
+
label="🎶 Generated Audio File",
|
132 |
+
type="filepath",
|
133 |
+
interactive=False
|
134 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
135 |
|
136 |
# Footer
|
137 |
gr.Markdown("""
|
138 |
<br><hr>
|
139 |
<p style="text-align: center; font-size: 0.9em;">
|
140 |
+
Created with ❤️ by <a href="https://bilsimaging.com" target="_blank">bilsimaging.com</a>
|
141 |
+
Special thanks to the <strong>Hugging Face community</strong> for their incredible support and tools!
|
142 |
</p>
|
143 |
""", elem_id="footer")
|
144 |
|