Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -1,31 +1,37 @@
|
|
1 |
import os
|
2 |
import shlex
|
3 |
import subprocess
|
4 |
-
import threading
|
5 |
import tempfile
|
6 |
import traceback
|
7 |
from pathlib import Path
|
8 |
|
|
|
9 |
os.system("pip install -r requirements.txt")
|
|
|
|
|
10 |
os.system("wget https://huggingface.co/stepfun-ai/Step-Audio-2-mini/resolve/main/token2wav/campplus.onnx -P token2wav")
|
11 |
os.system("wget https://huggingface.co/stepfun-ai/Step-Audio-2-mini/resolve/main/token2wav/flow.pt -P token2wav")
|
12 |
os.system("wget https://huggingface.co/stepfun-ai/Step-Audio-2-mini/resolve/main/token2wav/flow.yaml -P token2wav")
|
13 |
os.system("wget https://huggingface.co/stepfun-ai/Step-Audio-2-mini/resolve/main/token2wav/hift.pt -P token2wav")
|
14 |
|
15 |
-
#
|
16 |
hf_token = os.getenv("HF_TOKEN", None)
|
17 |
-
|
|
|
18 |
|
19 |
import spaces
|
20 |
import gradio as gr
|
21 |
|
22 |
-
def save_tmp_audio(audio_bytes, cache_dir):
|
|
|
23 |
os.makedirs(cache_dir, exist_ok=True)
|
24 |
with tempfile.NamedTemporaryFile(dir=cache_dir, delete=False, suffix=".wav") as temp_audio:
|
25 |
temp_audio.write(audio_bytes)
|
26 |
-
|
|
|
27 |
|
28 |
def add_message(chatbot, history, mic, text):
|
|
|
29 |
if not mic and not text:
|
30 |
return chatbot, history, "Input is empty"
|
31 |
|
@@ -36,49 +42,47 @@ def add_message(chatbot, history, mic, text):
|
|
36 |
chatbot.append({"role": "user", "content": {"path": mic}})
|
37 |
history.append({"role": "human", "content": [{"type": "audio", "audio": mic}]})
|
38 |
|
|
|
39 |
return chatbot, history, None
|
40 |
|
41 |
-
|
|
|
|
|
42 |
return [], [{"role": "system", "content": system_prompt}]
|
43 |
|
44 |
-
_AUDIO_MODEL = None
|
45 |
-
_TOKEN2WAV = None
|
46 |
-
_INIT_LOCK = threading.Lock()
|
47 |
|
48 |
-
|
|
|
|
|
|
|
49 |
"""
|
50 |
-
|
|
|
51 |
"""
|
52 |
-
global
|
53 |
-
if
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
_AUDIO_MODEL = StepAudio2(model_path)
|
61 |
-
_TOKEN2WAV = Token2wav(token2wav_dir)
|
62 |
-
|
63 |
-
return _AUDIO_MODEL, _TOKEN2WAV
|
64 |
|
65 |
@spaces.GPU
|
66 |
-
def predict(chatbot, history, prompt_wav, cache_dir, model_path
|
67 |
"""
|
68 |
-
|
69 |
-
Heavy
|
70 |
"""
|
71 |
try:
|
72 |
-
audio_model, token2wav =
|
73 |
|
74 |
-
# Stream start marker
|
75 |
history.append({
|
76 |
"role": "assistant",
|
77 |
"content": [{"type": "text", "text": "<tts_start>"}],
|
78 |
"eot": False
|
79 |
})
|
80 |
|
81 |
-
# Your original generation call
|
82 |
tokens, text, audio_tokens = audio_model(
|
83 |
history,
|
84 |
max_new_tokens=4096,
|
@@ -86,21 +90,23 @@ def predict(chatbot, history, prompt_wav, cache_dir, model_path, token2wav_dir):
|
|
86 |
repetition_penalty=1.05,
|
87 |
do_sample=True
|
88 |
)
|
|
|
89 |
|
90 |
-
# Convert tokens ->
|
91 |
audio_bytes = token2wav(audio_tokens, prompt_wav)
|
92 |
|
93 |
-
#
|
94 |
audio_path = save_tmp_audio(audio_bytes, cache_dir)
|
95 |
-
chatbot.append({"role": "assistant", "content": {"path": audio_path}})
|
96 |
|
97 |
-
#
|
|
|
98 |
history[-1]["content"].append({"type": "token", "token": tokens})
|
99 |
history[-1]["eot"] = True
|
100 |
|
101 |
except Exception:
|
102 |
print(traceback.format_exc())
|
103 |
-
gr.Warning("Some error
|
|
|
104 |
return chatbot, history
|
105 |
|
106 |
def _launch_demo(args):
|
@@ -112,71 +118,85 @@ def _launch_demo(args):
|
|
112 |
label="System Prompt",
|
113 |
value=(
|
114 |
"你的名字叫做小跃,是由阶跃星辰公司训练出来的语音大模型。\n"
|
115 |
-
"
|
|
|
116 |
"今天是2025年8月29日,星期五\n"
|
117 |
"请用默认女声与用户交流。"
|
118 |
),
|
119 |
-
lines=2
|
120 |
)
|
121 |
|
122 |
-
chatbot = gr.Chatbot(
|
123 |
-
|
124 |
-
|
|
|
|
|
125 |
|
126 |
-
#
|
127 |
-
|
128 |
-
text = gr.Textbox(placeholder="Enter message ...")
|
129 |
|
130 |
-
|
131 |
-
|
132 |
-
token2wav_dir = "token2wav"
|
133 |
-
prompt_wav = "assets/default_female.wav"
|
134 |
-
cache_dir = "/tmp/stepaudio2"
|
135 |
|
136 |
with gr.Row():
|
137 |
clean_btn = gr.Button("🧹 Clear History (清除历史)")
|
138 |
regen_btn = gr.Button("🤔️ Regenerate (重试)")
|
139 |
submit_btn = gr.Button("🚀 Submit")
|
140 |
|
141 |
-
|
142 |
-
|
143 |
-
chatbot, history, error = add_message(chatbot, history, mic, text)
|
144 |
if error:
|
145 |
gr.Warning(error)
|
146 |
-
return
|
147 |
-
|
148 |
-
|
|
|
|
|
|
|
|
|
|
|
149 |
|
150 |
submit_btn.click(
|
151 |
fn=on_submit,
|
152 |
-
inputs=[chatbot, history, mic, text
|
153 |
outputs=[chatbot, history, mic, text],
|
154 |
concurrency_limit=4,
|
155 |
concurrency_id="gpu_queue",
|
156 |
)
|
157 |
|
|
|
|
|
|
|
158 |
clean_btn.click(
|
159 |
-
fn=
|
160 |
inputs=[system_prompt],
|
161 |
outputs=[chatbot, history],
|
162 |
)
|
163 |
|
164 |
-
def
|
165 |
-
#
|
166 |
-
while
|
167 |
-
|
168 |
-
while
|
169 |
-
|
170 |
-
|
|
|
|
|
|
|
|
|
|
|
171 |
|
172 |
regen_btn.click(
|
173 |
-
fn=
|
174 |
-
inputs=[chatbot, history
|
175 |
outputs=[chatbot, history],
|
176 |
concurrency_id="gpu_queue",
|
177 |
)
|
178 |
|
179 |
-
|
|
|
|
|
|
|
180 |
|
181 |
if __name__ == "__main__":
|
182 |
from argparse import ArgumentParser
|
@@ -190,8 +210,4 @@ if __name__ == "__main__":
|
|
190 |
args = parser.parse_args()
|
191 |
|
192 |
os.environ["GRADIO_TEMP_DIR"] = args.cache_dir
|
193 |
-
os.makedirs(args.cache_dir, exist_ok=True)
|
194 |
-
|
195 |
-
# NOTE: Do NOT instantiate heavy models here.
|
196 |
-
# They will be created lazily inside predict() via _ensure_models(...).
|
197 |
_launch_demo(args)
|
|
|
1 |
import os
|
2 |
import shlex
|
3 |
import subprocess
|
|
|
4 |
import tempfile
|
5 |
import traceback
|
6 |
from pathlib import Path
|
7 |
|
8 |
+
# --- Install / fetch runtime deps & assets ---
|
9 |
os.system("pip install -r requirements.txt")
|
10 |
+
|
11 |
+
# Download token2wav assets
|
12 |
os.system("wget https://huggingface.co/stepfun-ai/Step-Audio-2-mini/resolve/main/token2wav/campplus.onnx -P token2wav")
|
13 |
os.system("wget https://huggingface.co/stepfun-ai/Step-Audio-2-mini/resolve/main/token2wav/flow.pt -P token2wav")
|
14 |
os.system("wget https://huggingface.co/stepfun-ai/Step-Audio-2-mini/resolve/main/token2wav/flow.yaml -P token2wav")
|
15 |
os.system("wget https://huggingface.co/stepfun-ai/Step-Audio-2-mini/resolve/main/token2wav/hift.pt -P token2wav")
|
16 |
|
17 |
+
# Hugging Face token (optional)
|
18 |
hf_token = os.getenv("HF_TOKEN", None)
|
19 |
+
if hf_token is not None:
|
20 |
+
os.environ["HF_TOKEN"] = hf_token
|
21 |
|
22 |
import spaces
|
23 |
import gradio as gr
|
24 |
|
25 |
+
def save_tmp_audio(audio_bytes: bytes, cache_dir: str) -> str:
|
26 |
+
"""Save raw wav bytes to a temporary file and return path."""
|
27 |
os.makedirs(cache_dir, exist_ok=True)
|
28 |
with tempfile.NamedTemporaryFile(dir=cache_dir, delete=False, suffix=".wav") as temp_audio:
|
29 |
temp_audio.write(audio_bytes)
|
30 |
+
return temp_audio.name
|
31 |
+
|
32 |
|
33 |
def add_message(chatbot, history, mic, text):
|
34 |
+
"""Append user text or audio to the chat + history."""
|
35 |
if not mic and not text:
|
36 |
return chatbot, history, "Input is empty"
|
37 |
|
|
|
42 |
chatbot.append({"role": "user", "content": {"path": mic}})
|
43 |
history.append({"role": "human", "content": [{"type": "audio", "audio": mic}]})
|
44 |
|
45 |
+
print(f"{history=}")
|
46 |
return chatbot, history, None
|
47 |
|
48 |
+
|
49 |
+
def reset_state(system_prompt: str):
|
50 |
+
"""Reset chat to a single system message."""
|
51 |
return [], [{"role": "system", "content": system_prompt}]
|
52 |
|
|
|
|
|
|
|
53 |
|
54 |
+
_MODEL = None
|
55 |
+
_TOK2WAV = None
|
56 |
+
|
57 |
+
def _get_models(model_path: str):
|
58 |
"""
|
59 |
+
Lazily load heavy, non-picklable models INSIDE the worker process
|
60 |
+
and cache them in module globals for reuse.
|
61 |
"""
|
62 |
+
global _MODEL, _TOK2WAV
|
63 |
+
if _MODEL is None or _TOK2WAV is None:
|
64 |
+
# Import here so the objects are constructed in the worker
|
65 |
+
from stepaudio2 import StepAudio2
|
66 |
+
from token2wav import Token2wav
|
67 |
+
_MODEL = StepAudio2(model_path)
|
68 |
+
_TOK2WAV = Token2wav("token2wav")
|
69 |
+
return _MODEL, _TOK2WAV
|
|
|
|
|
|
|
|
|
70 |
|
71 |
@spaces.GPU
|
72 |
+
def predict(chatbot, history, prompt_wav, cache_dir, model_path="Step-Audio-2-mini"):
|
73 |
"""
|
74 |
+
Run generation on GPU worker. All args must be picklable (strings, lists, dicts).
|
75 |
+
Heavy models are created via _get_models() inside this process.
|
76 |
"""
|
77 |
try:
|
78 |
+
audio_model, token2wav = _get_models(model_path)
|
79 |
|
|
|
80 |
history.append({
|
81 |
"role": "assistant",
|
82 |
"content": [{"type": "text", "text": "<tts_start>"}],
|
83 |
"eot": False
|
84 |
})
|
85 |
|
|
|
86 |
tokens, text, audio_tokens = audio_model(
|
87 |
history,
|
88 |
max_new_tokens=4096,
|
|
|
90 |
repetition_penalty=1.05,
|
91 |
do_sample=True
|
92 |
)
|
93 |
+
print(f"predict text={text!r}")
|
94 |
|
95 |
+
# Convert tokens -> waveform bytes using token2wav
|
96 |
audio_bytes = token2wav(audio_tokens, prompt_wav)
|
97 |
|
98 |
+
# Persist to temp .wav for the UI
|
99 |
audio_path = save_tmp_audio(audio_bytes, cache_dir)
|
|
|
100 |
|
101 |
+
# Append assistant audio message
|
102 |
+
chatbot.append({"role": "assistant", "content": {"path": audio_path}})
|
103 |
history[-1]["content"].append({"type": "token", "token": tokens})
|
104 |
history[-1]["eot"] = True
|
105 |
|
106 |
except Exception:
|
107 |
print(traceback.format_exc())
|
108 |
+
gr.Warning("Some error happend, please try again.")
|
109 |
+
|
110 |
return chatbot, history
|
111 |
|
112 |
def _launch_demo(args):
|
|
|
118 |
label="System Prompt",
|
119 |
value=(
|
120 |
"你的名字叫做小跃,是由阶跃星辰公司训练出来的语音大模型。\n"
|
121 |
+
"你情感细腻,观察能力强,擅长分析用户的内容,并作出善解人意的回复,"
|
122 |
+
"说话的过程中时刻注意用户的感受,富有同理心,提供多样的情绪价值。\n"
|
123 |
"今天是2025年8月29日,星期五\n"
|
124 |
"请用默认女声与用户交流。"
|
125 |
),
|
126 |
+
lines=2,
|
127 |
)
|
128 |
|
129 |
+
chatbot = gr.Chatbot(
|
130 |
+
elem_id="chatbot",
|
131 |
+
min_height=800,
|
132 |
+
type="messages",
|
133 |
+
)
|
134 |
|
135 |
+
# Initialize history with current system prompt value
|
136 |
+
history = gr.State([{"role": "system", "content": system_prompt.value}])
|
|
|
137 |
|
138 |
+
mic = gr.Audio(type="filepath", label="🎤 Speak (optional)")
|
139 |
+
text = gr.Textbox(placeholder="Enter message ...", label="💬 Text")
|
|
|
|
|
|
|
140 |
|
141 |
with gr.Row():
|
142 |
clean_btn = gr.Button("🧹 Clear History (清除历史)")
|
143 |
regen_btn = gr.Button("🤔️ Regenerate (重试)")
|
144 |
submit_btn = gr.Button("🚀 Submit")
|
145 |
|
146 |
+
def on_submit(chatbot_val, history_val, mic_val, text_val):
|
147 |
+
chatbot2, history2, error = add_message(chatbot_val, history_val, mic_val, text_val)
|
|
|
148 |
if error:
|
149 |
gr.Warning(error)
|
150 |
+
return chatbot2, history2, None, None
|
151 |
+
# Run GPU inference with only picklable args
|
152 |
+
chatbot2, history2 = predict(
|
153 |
+
chatbot2, history2,
|
154 |
+
args.prompt_wav, args.cache_dir,
|
155 |
+
model_path=args.model_path
|
156 |
+
)
|
157 |
+
return chatbot2, history2, None, None
|
158 |
|
159 |
submit_btn.click(
|
160 |
fn=on_submit,
|
161 |
+
inputs=[chatbot, history, mic, text],
|
162 |
outputs=[chatbot, history, mic, text],
|
163 |
concurrency_limit=4,
|
164 |
concurrency_id="gpu_queue",
|
165 |
)
|
166 |
|
167 |
+
def on_clean(system_prompt_text):
|
168 |
+
return reset_state(system_prompt_text)
|
169 |
+
|
170 |
clean_btn.click(
|
171 |
+
fn=on_clean,
|
172 |
inputs=[system_prompt],
|
173 |
outputs=[chatbot, history],
|
174 |
)
|
175 |
|
176 |
+
def on_regenerate(chatbot_val, history_val):
|
177 |
+
# Drop last assistant turn(s) to regenerate
|
178 |
+
while chatbot_val and chatbot_val[-1]["role"] == "assistant":
|
179 |
+
chatbot_val.pop()
|
180 |
+
while history_val and history_val[-1]["role"] == "assistant":
|
181 |
+
print(f"discard {history_val[-1]}")
|
182 |
+
history_val.pop()
|
183 |
+
return predict(
|
184 |
+
chatbot_val, history_val,
|
185 |
+
args.prompt_wav, args.cache_dir,
|
186 |
+
model_path=args.model_path
|
187 |
+
)
|
188 |
|
189 |
regen_btn.click(
|
190 |
+
fn=on_regenerate,
|
191 |
+
inputs=[chatbot, history],
|
192 |
outputs=[chatbot, history],
|
193 |
concurrency_id="gpu_queue",
|
194 |
)
|
195 |
|
196 |
+
demo.queue().launch(
|
197 |
+
server_port=args.server_port,
|
198 |
+
server_name=args.server_name,
|
199 |
+
)
|
200 |
|
201 |
if __name__ == "__main__":
|
202 |
from argparse import ArgumentParser
|
|
|
210 |
args = parser.parse_args()
|
211 |
|
212 |
os.environ["GRADIO_TEMP_DIR"] = args.cache_dir
|
|
|
|
|
|
|
|
|
213 |
_launch_demo(args)
|