Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -47,29 +47,6 @@ MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
|
|
47 |
|
48 |
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
49 |
|
50 |
-
# Define a helper function that returns HTML for a progress bar with a label.
|
51 |
-
def progress_bar_html_with_label(label="Thinking..."):
|
52 |
-
return f"""
|
53 |
-
<div style="font-weight: bold; margin-bottom: 5px;">{label}</div>
|
54 |
-
<div id="progress-container" style="width: 100%; background-color: #eee; border-radius: 4px; overflow: hidden;">
|
55 |
-
<div id="progress-bar" style="width: 0%; height: 10px; background-color: limegreen; transition: width 0.1s;"></div>
|
56 |
-
</div>
|
57 |
-
<script>
|
58 |
-
(function() {{
|
59 |
-
let progressBar = document.getElementById("progress-bar");
|
60 |
-
let width = 0;
|
61 |
-
let interval = setInterval(function(){{
|
62 |
-
if(width < 100) {{
|
63 |
-
width += 1;
|
64 |
-
progressBar.style.width = width + "%";
|
65 |
-
}} else {{
|
66 |
-
clearInterval(interval);
|
67 |
-
}}
|
68 |
-
}}, 100);
|
69 |
-
}})();
|
70 |
-
</script>
|
71 |
-
"""
|
72 |
-
|
73 |
# Load text-only model and tokenizer
|
74 |
model_id = "prithivMLmods/FastThink-0.5B-Tiny"
|
75 |
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
@@ -110,6 +87,22 @@ def clean_chat_history(chat_history):
|
|
110 |
cleaned.append(msg)
|
111 |
return cleaned
|
112 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
113 |
# Environment variables and parameters for Stable Diffusion XL
|
114 |
MODEL_ID_SD = os.getenv("MODEL_VAL_PATH") # SDXL Model repository path via env variable
|
115 |
MAX_IMAGE_SIZE = int(os.getenv("MAX_IMAGE_SIZE", "4096"))
|
@@ -214,29 +207,44 @@ def generate(
|
|
214 |
Special commands:
|
215 |
- "@tts1" or "@tts2": triggers text-to-speech.
|
216 |
- "@image": triggers image generation using the SDXL pipeline.
|
|
|
|
|
|
|
217 |
"""
|
218 |
text = input_dict["text"]
|
219 |
files = input_dict.get("files", [])
|
220 |
|
221 |
-
#
|
222 |
if text.strip().lower().startswith("@image"):
|
223 |
prompt = text[len("@image"):].strip()
|
224 |
-
#
|
225 |
-
|
226 |
-
|
227 |
-
|
228 |
-
|
229 |
-
|
230 |
-
|
231 |
-
|
232 |
-
|
233 |
-
|
234 |
-
|
235 |
-
|
236 |
-
|
237 |
-
|
238 |
-
|
239 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
240 |
yield gr.Image(image_paths[0])
|
241 |
return # Exit early
|
242 |
|
@@ -247,14 +255,16 @@ def generate(
|
|
247 |
if is_tts and voice_index:
|
248 |
voice = TTS_VOICES[voice_index - 1]
|
249 |
text = text.replace(f"{tts_prefix}{voice_index}", "").strip()
|
|
|
250 |
conversation = [{"role": "user", "content": text}]
|
251 |
else:
|
252 |
voice = None
|
|
|
253 |
text = text.replace(tts_prefix, "").strip()
|
254 |
conversation = clean_chat_history(chat_history)
|
255 |
conversation.append({"role": "user", "content": text})
|
256 |
|
257 |
-
# Multimodal
|
258 |
if files:
|
259 |
if len(files) > 1:
|
260 |
images = [load_image(image) for image in files]
|
@@ -277,21 +287,20 @@ def generate(
|
|
277 |
thread.start()
|
278 |
|
279 |
buffer = ""
|
280 |
-
|
281 |
-
|
|
|
282 |
for new_text in streamer:
|
283 |
buffer += new_text
|
284 |
buffer = buffer.replace("<|im_end|>", "")
|
285 |
-
|
286 |
-
|
287 |
-
|
288 |
-
|
289 |
-
|
290 |
-
yield gr.HTML(html)
|
291 |
-
# Final output: only the generated text without the progress bar.
|
292 |
yield buffer
|
293 |
else:
|
294 |
-
# Text-only generation
|
295 |
input_ids = tokenizer.apply_chat_template(conversation, add_generation_prompt=True, return_tensors="pt")
|
296 |
if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
|
297 |
input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
|
@@ -313,19 +322,18 @@ def generate(
|
|
313 |
t.start()
|
314 |
|
315 |
outputs = []
|
316 |
-
|
317 |
-
#
|
318 |
-
yield
|
319 |
for new_text in streamer:
|
320 |
outputs.append(new_text)
|
321 |
-
|
322 |
-
|
323 |
-
|
324 |
-
|
325 |
-
|
326 |
-
|
327 |
-
|
328 |
-
# Final output: just the final text.
|
329 |
yield final_response
|
330 |
|
331 |
# If TTS was requested, convert the final response to speech.
|
|
|
47 |
|
48 |
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
49 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
50 |
# Load text-only model and tokenizer
|
51 |
model_id = "prithivMLmods/FastThink-0.5B-Tiny"
|
52 |
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
|
|
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"))
|
|
|
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 |
+
# Image generation branch
|
218 |
if text.strip().lower().startswith("@image"):
|
219 |
prompt = text[len("@image"):].strip()
|
220 |
+
# Use a container to capture the result from the thread.
|
221 |
+
result_container = []
|
222 |
+
def run_image():
|
223 |
+
result_container.append(generate_image_fn(
|
224 |
+
prompt=prompt,
|
225 |
+
negative_prompt="",
|
226 |
+
use_negative_prompt=False,
|
227 |
+
seed=1,
|
228 |
+
width=1024,
|
229 |
+
height=1024,
|
230 |
+
guidance_scale=3,
|
231 |
+
num_inference_steps=25,
|
232 |
+
randomize_seed=True,
|
233 |
+
use_resolution_binning=True,
|
234 |
+
num_images=1,
|
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 |
|
|
|
255 |
if is_tts and voice_index:
|
256 |
voice = TTS_VOICES[voice_index - 1]
|
257 |
text = text.replace(f"{tts_prefix}{voice_index}", "").strip()
|
258 |
+
# Clear previous chat history for a fresh TTS request.
|
259 |
conversation = [{"role": "user", "content": text}]
|
260 |
else:
|
261 |
voice = None
|
262 |
+
# Remove any stray @tts tags and build the conversation history.
|
263 |
text = text.replace(tts_prefix, "").strip()
|
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 |
thread.start()
|
288 |
|
289 |
buffer = ""
|
290 |
+
start_time = time.time()
|
291 |
+
# Initial progress bar for multimodal inference.
|
292 |
+
yield render_progress_bar("Thinking...", 0)
|
293 |
for new_text in streamer:
|
294 |
buffer += new_text
|
295 |
buffer = buffer.replace("<|im_end|>", "")
|
296 |
+
progress = min(95, int((time.time() - start_time) / 5 * 95))
|
297 |
+
yield render_progress_bar("Thinking...", progress, output_text=buffer)
|
298 |
+
# Final progress update (100%).
|
299 |
+
yield render_progress_bar("Thinking...", 100, output_text=buffer)
|
300 |
+
# Then yield final response (progress bar update no longer shown).
|
|
|
|
|
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:]
|
|
|
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 |
+
progress = min(95, int((time.time() - start_time) / 5 * 95))
|
332 |
+
yield render_progress_bar("Thinking...", progress, output_text=current_text)
|
333 |
+
final_response = "".join(outputs)
|
334 |
+
# Final update (100% progress).
|
335 |
+
yield render_progress_bar("Thinking...", 100, output_text=final_response)
|
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.
|