Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -23,6 +23,7 @@ from transformers import (
|
|
23 |
from transformers.image_utils import load_image
|
24 |
from diffusers import StableDiffusionXLPipeline, EulerAncestralDiscreteScheduler
|
25 |
|
|
|
26 |
DESCRIPTION = """
|
27 |
# QwQ Edge 💬
|
28 |
"""
|
@@ -41,6 +42,23 @@ h1 {
|
|
41 |
}
|
42 |
'''
|
43 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
44 |
MAX_MAX_NEW_TOKENS = 2048
|
45 |
DEFAULT_MAX_NEW_TOKENS = 1024
|
46 |
MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
|
@@ -87,22 +105,6 @@ def clean_chat_history(chat_history):
|
|
87 |
cleaned.append(msg)
|
88 |
return cleaned
|
89 |
|
90 |
-
# A helper function to render a progress bar using HTML.
|
91 |
-
def render_progress_bar(label: str, progress: int, output_text: str = "") -> str:
|
92 |
-
"""
|
93 |
-
Returns an HTML snippet containing a label, a progress bar (red background with a green inner bar),
|
94 |
-
and optionally some output text.
|
95 |
-
"""
|
96 |
-
return f'''
|
97 |
-
<div style="margin-bottom: 10px;">
|
98 |
-
<div style="font-weight: bold; margin-bottom: 5px;">{label}</div>
|
99 |
-
<div style="width: 100%; background-color: red; border-radius: 5px; overflow: hidden; height: 10px;">
|
100 |
-
<div style="width: {progress}%; background-color: green; height: 100%; transition: width 0.3s;"></div>
|
101 |
-
</div>
|
102 |
-
<div style="margin-top: 10px;">{output_text}</div>
|
103 |
-
</div>
|
104 |
-
'''
|
105 |
-
|
106 |
# Environment variables and parameters for Stable Diffusion XL
|
107 |
MODEL_ID_SD = os.getenv("MODEL_VAL_PATH") # SDXL Model repository path via env variable
|
108 |
MAX_IMAGE_SIZE = int(os.getenv("MAX_IMAGE_SIZE", "4096"))
|
@@ -183,6 +185,7 @@ def generate_image_fn(
|
|
183 |
batch_options["prompt"] = options["prompt"][i:i+BATCH_SIZE]
|
184 |
if "negative_prompt" in batch_options and batch_options["negative_prompt"] is not None:
|
185 |
batch_options["negative_prompt"] = options["negative_prompt"][i:i+BATCH_SIZE]
|
|
|
186 |
if device.type == "cuda":
|
187 |
with torch.autocast("cuda", dtype=torch.float16):
|
188 |
outputs = sd_pipe(**batch_options)
|
@@ -207,51 +210,36 @@ def generate(
|
|
207 |
Special commands:
|
208 |
- "@tts1" or "@tts2": triggers text-to-speech.
|
209 |
- "@image": triggers image generation using the SDXL pipeline.
|
210 |
-
|
211 |
-
Instead of yielding a simple "Thinking..." text, an animated progress bar is shown (via an HTML snippet)
|
212 |
-
that goes from red to green. When the inference is complete the progress bar is replaced by the final result.
|
213 |
"""
|
214 |
text = input_dict["text"]
|
215 |
files = input_dict.get("files", [])
|
216 |
|
217 |
-
|
|
|
|
|
|
|
218 |
if text.strip().lower().startswith("@image"):
|
|
|
219 |
prompt = text[len("@image"):].strip()
|
220 |
-
#
|
221 |
-
|
222 |
-
|
223 |
-
|
224 |
-
|
225 |
-
|
226 |
-
|
227 |
-
|
228 |
-
|
229 |
-
|
230 |
-
|
231 |
-
|
232 |
-
|
233 |
-
|
234 |
-
|
235 |
-
|
236 |
-
thread = Thread(target=run_image)
|
237 |
-
thread.start()
|
238 |
-
start_time = time.time()
|
239 |
-
# Simulate progress bar updates while image generation is running.
|
240 |
-
while thread.is_alive():
|
241 |
-
progress = min(95, int((time.time() - start_time) / 5 * 95))
|
242 |
-
yield render_progress_bar("Generating Image", progress)
|
243 |
-
time.sleep(0.5)
|
244 |
-
thread.join()
|
245 |
-
# Final update before showing the result.
|
246 |
-
yield render_progress_bar("Generating Image", 100)
|
247 |
-
image_paths, used_seed = result_container[0]
|
248 |
yield gr.Image(image_paths[0])
|
249 |
return # Exit early
|
250 |
|
251 |
-
tts_prefix = "@tts"
|
252 |
-
is_tts = any(text.strip().lower().startswith(f"{tts_prefix}{i}") for i in range(1, 3))
|
253 |
-
voice_index = next((i for i in range(1, 3) if text.strip().lower().startswith(f"{tts_prefix}{i}")), None)
|
254 |
-
|
255 |
if is_tts and voice_index:
|
256 |
voice = TTS_VOICES[voice_index - 1]
|
257 |
text = text.replace(f"{tts_prefix}{voice_index}", "").strip()
|
@@ -264,7 +252,6 @@ def generate(
|
|
264 |
conversation = clean_chat_history(chat_history)
|
265 |
conversation.append({"role": "user", "content": text})
|
266 |
|
267 |
-
# Multimodal (image + text) branch
|
268 |
if files:
|
269 |
if len(files) > 1:
|
270 |
images = [load_image(image) for image in files]
|
@@ -287,20 +274,17 @@ def generate(
|
|
287 |
thread.start()
|
288 |
|
289 |
buffer = ""
|
290 |
-
|
291 |
-
|
292 |
-
yield render_progress_bar("Thinking...", 0)
|
293 |
for new_text in streamer:
|
294 |
buffer += new_text
|
295 |
buffer = buffer.replace("<|im_end|>", "")
|
296 |
-
|
297 |
-
|
298 |
-
|
299 |
-
|
300 |
-
|
301 |
-
yield buffer
|
302 |
else:
|
303 |
-
# Text-only generation branch.
|
304 |
input_ids = tokenizer.apply_chat_template(conversation, add_generation_prompt=True, return_tensors="pt")
|
305 |
if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
|
306 |
input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
|
@@ -321,20 +305,18 @@ def generate(
|
|
321 |
t = Thread(target=model.generate, kwargs=generation_kwargs)
|
322 |
t.start()
|
323 |
|
|
|
|
|
324 |
outputs = []
|
325 |
-
start_time = time.time()
|
326 |
-
# Initial progress bar update.
|
327 |
-
yield render_progress_bar("Thinking...", 0)
|
328 |
for new_text in streamer:
|
329 |
outputs.append(new_text)
|
330 |
current_text = "".join(outputs)
|
331 |
-
|
332 |
-
|
|
|
333 |
final_response = "".join(outputs)
|
334 |
-
# Final
|
335 |
-
yield
|
336 |
-
# Finally, yield the final plain response so the progress bar disappears.
|
337 |
-
yield final_response
|
338 |
|
339 |
# If TTS was requested, convert the final response to speech.
|
340 |
if is_tts and voice:
|
|
|
23 |
from transformers.image_utils import load_image
|
24 |
from diffusers import StableDiffusionXLPipeline, EulerAncestralDiscreteScheduler
|
25 |
|
26 |
+
|
27 |
DESCRIPTION = """
|
28 |
# QwQ Edge 💬
|
29 |
"""
|
|
|
42 |
}
|
43 |
'''
|
44 |
|
45 |
+
def progress_bar_html(label: str) -> str:
|
46 |
+
"""Return an HTML snippet with a label and an animated, thin light-blue progress bar."""
|
47 |
+
return f"""
|
48 |
+
<div style="display: flex; align-items: center;">
|
49 |
+
<span style="margin-right: 8px;">{label}</span>
|
50 |
+
<div style="position: relative; width: 110px; height: 5px; background: #e0e0e0; border-radius: 5px; overflow: hidden;">
|
51 |
+
<div style="width: 100%; height: 100%; background-color: lightblue; animation: progress-bar-animation 1s linear infinite;"></div>
|
52 |
+
</div>
|
53 |
+
</div>
|
54 |
+
<style>
|
55 |
+
@keyframes progress-bar-animation {{
|
56 |
+
0% {{ transform: translateX(-100%); }}
|
57 |
+
100% {{ transform: translateX(100%); }}
|
58 |
+
}}
|
59 |
+
</style>
|
60 |
+
"""
|
61 |
+
|
62 |
MAX_MAX_NEW_TOKENS = 2048
|
63 |
DEFAULT_MAX_NEW_TOKENS = 1024
|
64 |
MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
|
|
|
105 |
cleaned.append(msg)
|
106 |
return cleaned
|
107 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
108 |
# Environment variables and parameters for Stable Diffusion XL
|
109 |
MODEL_ID_SD = os.getenv("MODEL_VAL_PATH") # SDXL Model repository path via env variable
|
110 |
MAX_IMAGE_SIZE = int(os.getenv("MAX_IMAGE_SIZE", "4096"))
|
|
|
185 |
batch_options["prompt"] = options["prompt"][i:i+BATCH_SIZE]
|
186 |
if "negative_prompt" in batch_options and batch_options["negative_prompt"] is not None:
|
187 |
batch_options["negative_prompt"] = options["negative_prompt"][i:i+BATCH_SIZE]
|
188 |
+
# Wrap the pipeline call in autocast if using CUDA
|
189 |
if device.type == "cuda":
|
190 |
with torch.autocast("cuda", dtype=torch.float16):
|
191 |
outputs = sd_pipe(**batch_options)
|
|
|
210 |
Special commands:
|
211 |
- "@tts1" or "@tts2": triggers text-to-speech.
|
212 |
- "@image": triggers image generation using the SDXL pipeline.
|
|
|
|
|
|
|
213 |
"""
|
214 |
text = input_dict["text"]
|
215 |
files = input_dict.get("files", [])
|
216 |
|
217 |
+
tts_prefix = "@tts"
|
218 |
+
is_tts = any(text.strip().lower().startswith(f"{tts_prefix}{i}") for i in range(1, 3))
|
219 |
+
voice_index = next((i for i in range(1, 3) if text.strip().lower().startswith(f"{tts_prefix}{i}")), None)
|
220 |
+
|
221 |
if text.strip().lower().startswith("@image"):
|
222 |
+
# Remove the "@image" tag and use the rest as prompt
|
223 |
prompt = text[len("@image"):].strip()
|
224 |
+
# Yield progress bar for image generation
|
225 |
+
yield progress_bar_html("Generating Image")
|
226 |
+
image_paths, used_seed = generate_image_fn(
|
227 |
+
prompt=prompt,
|
228 |
+
negative_prompt="",
|
229 |
+
use_negative_prompt=False,
|
230 |
+
seed=1,
|
231 |
+
width=1024,
|
232 |
+
height=1024,
|
233 |
+
guidance_scale=3,
|
234 |
+
num_inference_steps=25,
|
235 |
+
randomize_seed=True,
|
236 |
+
use_resolution_binning=True,
|
237 |
+
num_images=1,
|
238 |
+
)
|
239 |
+
# Yield the generated image, replacing the progress bar
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
240 |
yield gr.Image(image_paths[0])
|
241 |
return # Exit early
|
242 |
|
|
|
|
|
|
|
|
|
243 |
if is_tts and voice_index:
|
244 |
voice = TTS_VOICES[voice_index - 1]
|
245 |
text = text.replace(f"{tts_prefix}{voice_index}", "").strip()
|
|
|
252 |
conversation = clean_chat_history(chat_history)
|
253 |
conversation.append({"role": "user", "content": text})
|
254 |
|
|
|
255 |
if files:
|
256 |
if len(files) > 1:
|
257 |
images = [load_image(image) for image in files]
|
|
|
274 |
thread.start()
|
275 |
|
276 |
buffer = ""
|
277 |
+
# Yield initial progress bar for multimodal generation
|
278 |
+
yield progress_bar_html("Thinking...")
|
|
|
279 |
for new_text in streamer:
|
280 |
buffer += new_text
|
281 |
buffer = buffer.replace("<|im_end|>", "")
|
282 |
+
time.sleep(0.01)
|
283 |
+
# Update with partial text and progress bar
|
284 |
+
yield f"<div>{buffer}</div><div>{progress_bar_html('Thinking...')}</div>"
|
285 |
+
# Final output: remove progress bar
|
286 |
+
yield f"<div>{buffer}</div>"
|
|
|
287 |
else:
|
|
|
288 |
input_ids = tokenizer.apply_chat_template(conversation, add_generation_prompt=True, return_tensors="pt")
|
289 |
if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
|
290 |
input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
|
|
|
305 |
t = Thread(target=model.generate, kwargs=generation_kwargs)
|
306 |
t.start()
|
307 |
|
308 |
+
# Yield initial progress bar for text generation
|
309 |
+
yield progress_bar_html("Thinking...")
|
310 |
outputs = []
|
|
|
|
|
|
|
311 |
for new_text in streamer:
|
312 |
outputs.append(new_text)
|
313 |
current_text = "".join(outputs)
|
314 |
+
time.sleep(0.01)
|
315 |
+
# Update message with partial text and progress bar
|
316 |
+
yield f"<div>{current_text}</div><div>{progress_bar_html('Thinking...')}</div>"
|
317 |
final_response = "".join(outputs)
|
318 |
+
# Final output: only the final response text, progress bar removed.
|
319 |
+
yield f"<div>{final_response}</div>"
|
|
|
|
|
320 |
|
321 |
# If TTS was requested, convert the final response to speech.
|
322 |
if is_tts and voice:
|