chong.zhang commited on
Commit
84d3675
·
1 Parent(s): 2237355
Files changed (1) hide show
  1. app.py +12 -27
app.py CHANGED
@@ -12,12 +12,7 @@ os.system('nvidia-smi')
12
  print(torch.backends.cudnn.version())
13
 
14
  def generate_filename():
15
- now = datetime.datetime.now()
16
- seconds_since_epoch = int(now.timestamp())
17
- # Convert seconds to string
18
- seconds_str = str(seconds_since_epoch)
19
- # Hash the string using SHA-256
20
- hash_object = hashlib.sha256(seconds_str.encode())
21
  hash_string = hash_object.hexdigest()
22
  return hash_string
23
 
@@ -60,6 +55,15 @@ def get_args(
60
  print(args)
61
  return args
62
 
 
 
 
 
 
 
 
 
 
63
  @spaces.GPU
64
  def music_generation(args):
65
  set_env_variables()
@@ -90,24 +94,6 @@ def music_generation(args):
90
  trim=args["trim"])
91
  return output_path
92
 
93
- def update_text():
94
- global text_input # Declare as global to modify the outer scope variable
95
- text_input = "New value set by button click"
96
- return text_input
97
-
98
- default_prompts = [
99
- "Experience soothing and sensual instrumental jazz with a touch of Bossa Nova, perfect for a relaxing restaurant or spa ambiance.",
100
- "Compose an uplifting R&B song.",
101
- "Create an emotional, introspective folk song with acoustic guitar and soft vocals."
102
- ]
103
-
104
- def trim_audio(audio_file, cut_seconds=5):
105
- audio, sr = torchaudio.load(audio_file)
106
- num_samples = cut_seconds * sr
107
- cutted_audio = audio[:, :num_samples]
108
- output_path = os.path.join(os.getcwd(), "audio_prompt_" + generate_filename() + ".wav")
109
- torchaudio.save(output_path, cutted_audio, sr)
110
- return output_path
111
 
112
  @spaces.GPU
113
  def demo_inspiremusic_t2m(text, model_name, chorus,
@@ -151,7 +137,6 @@ def main():
151
  value=30)
152
 
153
  with gr.Row(equal_height=True):
154
- # Textbox for custom input
155
  text_input = gr.Textbox(label="Input Text (For Text-to-Music Task)", value="Experience soothing and sensual instrumental jazz with a touch of Bossa Nova, perfect for a relaxing restaurant or spa ambiance.")
156
 
157
  audio_input = gr.Audio(label="Input Audio Prompt (For Music Continuation Task)",
@@ -177,8 +162,8 @@ def main():
177
  demo.launch()
178
 
179
  if __name__ == '__main__':
180
- model_list = ["InspireMusic-Base", "InspireMusic-1.5B-Long"]
181
- # model_list = ["InspireMusic-Base"]
182
  for model_name in model_list:
183
  model_dir = f"pretrained_models/{model_name}"
184
  if not os.path.isdir(model_dir):
 
12
  print(torch.backends.cudnn.version())
13
 
14
  def generate_filename():
15
+ hash_object = hashlib.sha256(str(int(datetime.datetime.now().timestamp())).encode())
 
 
 
 
 
16
  hash_string = hash_object.hexdigest()
17
  return hash_string
18
 
 
55
  print(args)
56
  return args
57
 
58
+
59
+ def trim_audio(audio_file, cut_seconds=5):
60
+ audio, sr = torchaudio.load(audio_file)
61
+ num_samples = cut_seconds * sr
62
+ cutted_audio = audio[:, :num_samples]
63
+ output_path = os.path.join(os.getcwd(), "audio_prompt_" + generate_filename() + ".wav")
64
+ torchaudio.save(output_path, cutted_audio, sr)
65
+ return output_path
66
+
67
  @spaces.GPU
68
  def music_generation(args):
69
  set_env_variables()
 
94
  trim=args["trim"])
95
  return output_path
96
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
97
 
98
  @spaces.GPU
99
  def demo_inspiremusic_t2m(text, model_name, chorus,
 
137
  value=30)
138
 
139
  with gr.Row(equal_height=True):
 
140
  text_input = gr.Textbox(label="Input Text (For Text-to-Music Task)", value="Experience soothing and sensual instrumental jazz with a touch of Bossa Nova, perfect for a relaxing restaurant or spa ambiance.")
141
 
142
  audio_input = gr.Audio(label="Input Audio Prompt (For Music Continuation Task)",
 
162
  demo.launch()
163
 
164
  if __name__ == '__main__':
165
+ model_list = ["InspireMusic-1.5B-Long", "InspireMusic-1.5B", "InspireMusic-Base"]
166
+
167
  for model_name in model_list:
168
  model_dir = f"pretrained_models/{model_name}"
169
  if not os.path.isdir(model_dir):