Spaces:
Running
on
Zero
Running
on
Zero
玙珲
commited on
Commit
·
e0ca852
1
Parent(s):
169061d
add thinking budget
Browse files
app.py
CHANGED
@@ -27,6 +27,25 @@ streamer = None
|
|
27 |
# This should point to the directory containing your SVG file.
|
28 |
CUR_DIR = os.path.dirname(os.path.abspath(__file__))
|
29 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
30 |
def submit_chat(chatbot, text_input):
|
31 |
response = ''
|
32 |
chatbot.append([text_input, response])
|
@@ -114,6 +133,8 @@ def run_inference(
|
|
114 |
do_sample: bool,
|
115 |
max_new_tokens: int,
|
116 |
enable_thinking: bool,
|
|
|
|
|
117 |
):
|
118 |
"""
|
119 |
Runs a single turn of inference and yields the output stream for a gr.Chatbot.
|
@@ -122,14 +143,11 @@ def run_inference(
|
|
122 |
prompt = chatbot[-1][0]
|
123 |
if (not image_input and not video_input and not prompt) or not prompt:
|
124 |
gr.Warning("A text prompt is required for generation.")
|
|
|
125 |
# MODIFICATION: Yield the current state and return to avoid errors
|
126 |
yield chatbot
|
127 |
return
|
128 |
|
129 |
-
# MODIFICATION: Append the new prompt to the existing history
|
130 |
-
# chatbot.append([prompt, ""])
|
131 |
-
# yield chatbot, "" # Yield the updated chat to show the user's prompt immediately
|
132 |
-
|
133 |
content = []
|
134 |
if image_input:
|
135 |
content.append({"type": "image", "image": image_input})
|
@@ -139,7 +157,7 @@ def run_inference(
|
|
139 |
content.append({"type": "video", "video": frames})
|
140 |
else:
|
141 |
gr.Warning("Failed to process the video file.")
|
142 |
-
chatbot
|
143 |
yield chatbot
|
144 |
return
|
145 |
|
@@ -154,7 +172,8 @@ def run_inference(
|
|
154 |
else:
|
155 |
input_ids, pixel_values, grid_thws = model.preprocess_inputs(messages=messages, add_generation_prompt=True, enable_thinking=enable_thinking)
|
156 |
except Exception as e:
|
157 |
-
|
|
|
158 |
yield chatbot
|
159 |
return
|
160 |
|
@@ -170,7 +189,10 @@ def run_inference(
|
|
170 |
"eos_token_id": model.text_tokenizer.eos_token_id,
|
171 |
"pad_token_id": model.text_tokenizer.pad_token_id,
|
172 |
"streamer": streamer,
|
173 |
-
"use_cache": True
|
|
|
|
|
|
|
174 |
}
|
175 |
|
176 |
with torch.inference_mode():
|
@@ -197,16 +219,11 @@ def run_inference(
|
|
197 |
chatbot[-1][1] = formatted_response
|
198 |
yield chatbot # Yield the final, formatted response
|
199 |
|
200 |
-
logger.info("[OVIS_CONV_START]")
|
201 |
-
[print(f'Q{i}:\n {request}\nA{i}:\n {answer}') for i, (request, answer) in enumerate(chatbot, 1)]
|
202 |
-
# print('New_Q:\n', text_input)
|
203 |
-
# print('New_A:\n', response)
|
204 |
logger.info("[OVIS_CONV_END]")
|
205 |
|
206 |
|
207 |
-
def clear_chat():
|
208 |
-
return [], None, ""
|
209 |
-
|
210 |
# --- UI Helper Functions ---
|
211 |
def toggle_media_input(choice: str) -> Tuple:
|
212 |
"""Switches visibility between Image/Video inputs and their corresponding examples."""
|
@@ -217,7 +234,6 @@ def toggle_media_input(choice: str) -> Tuple:
|
|
217 |
|
218 |
|
219 |
# --- Build Gradio Application ---
|
220 |
-
# @spaces.GPU
|
221 |
def build_demo(model_path: str):
|
222 |
"""Builds the Gradio user interface for the model."""
|
223 |
global model, streamer
|
@@ -231,7 +247,7 @@ def build_demo(model_path: str):
|
|
231 |
).to(device).eval()
|
232 |
|
233 |
text_tokenizer = model.text_tokenizer
|
234 |
-
streamer =
|
235 |
|
236 |
print("Model loaded successfully.")
|
237 |
|
@@ -257,10 +273,22 @@ def build_demo(model_path: str):
|
|
257 |
<center><font size=3><b>Ovis</b> has been open-sourced on <a href='https://huggingface.co/{model_path}'>😊 Huggingface</a> and <a href='https://github.com/AIDC-AI/Ovis'>🌟 GitHub</a>. If you find Ovis useful, a like❤️ or a star🌟 would be appreciated.</font></center>
|
258 |
"""
|
259 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
260 |
prompt_input = gr.Textbox(label="Prompt", placeholder="Enter your text here and press ENTER", lines=1, container=False)
|
261 |
with gr.Blocks(theme=gr.themes.Ocean()) as demo:
|
262 |
gr.HTML(html_header)
|
263 |
-
gr.Markdown("Note:
|
264 |
|
265 |
with gr.Row():
|
266 |
with gr.Column(scale=4):
|
@@ -270,10 +298,10 @@ def build_demo(model_path: str):
|
|
270 |
|
271 |
with gr.Accordion("Generation Settings", open=True):
|
272 |
do_sample = gr.Checkbox(label="Enable Sampling (Do Sample)", value=True)
|
273 |
-
|
274 |
-
|
275 |
-
|
276 |
-
|
277 |
|
278 |
with gr.Column(visible=True) as image_examples_col:
|
279 |
gr.Examples(
|
@@ -297,30 +325,50 @@ def build_demo(model_path: str):
|
|
297 |
generate_btn = gr.Button("Send", variant="primary")
|
298 |
clear_btn = gr.Button("Clear", variant="secondary")
|
299 |
|
|
|
|
|
300 |
input_type_radio.change(
|
301 |
fn=toggle_media_input,
|
302 |
inputs=input_type_radio,
|
303 |
outputs=[image_input, video_input, image_examples_col, video_examples_col]
|
304 |
)
|
305 |
|
306 |
-
#
|
307 |
-
|
308 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
309 |
|
310 |
generat_click_event = generate_btn.click(submit_chat, [chatbot, prompt_input], [chatbot, prompt_input]).then(run_inference, run_inputs, chatbot)
|
311 |
submit_event = prompt_input.submit(submit_chat, [chatbot, prompt_input], [chatbot, prompt_input]).then(run_inference, run_inputs, chatbot)
|
312 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
313 |
clear_btn.click(
|
314 |
-
fn=lambda: (
|
315 |
-
outputs=[chatbot, image_input, video_input, prompt_input
|
316 |
-
).then(
|
317 |
-
fn=toggle_media_input,
|
318 |
-
inputs=input_type_radio,
|
319 |
-
outputs=[image_input, video_input, image_examples_col, video_examples_col]
|
320 |
)
|
|
|
321 |
|
322 |
return demo
|
323 |
|
|
|
324 |
# --- Main Execution Block ---
|
325 |
# def parse_args():
|
326 |
# parser = argparse.ArgumentParser(description="Gradio interface for a single Multimodal Large Language Model.")
|
|
|
27 |
# This should point to the directory containing your SVG file.
|
28 |
CUR_DIR = os.path.dirname(os.path.abspath(__file__))
|
29 |
|
30 |
+
|
31 |
+
class MyTextIteratorStreamer(TextIteratorStreamer):
|
32 |
+
def manual_end(self):
|
33 |
+
"""Flushes any remaining cache and prints a newline to stdout."""
|
34 |
+
# Flush the cache, if it exists
|
35 |
+
if len(self.token_cache) > 0:
|
36 |
+
text = self.tokenizer.decode(self.token_cache, **self.decode_kwargs)
|
37 |
+
printable_text = text[self.print_len :]
|
38 |
+
self.token_cache = []
|
39 |
+
self.print_len = 0
|
40 |
+
else:
|
41 |
+
printable_text = ""
|
42 |
+
|
43 |
+
self.next_tokens_are_prompt = True
|
44 |
+
self.on_finalized_text(printable_text, stream_end=True)
|
45 |
+
|
46 |
+
def end(self):
|
47 |
+
pass
|
48 |
+
|
49 |
def submit_chat(chatbot, text_input):
|
50 |
response = ''
|
51 |
chatbot.append([text_input, response])
|
|
|
133 |
do_sample: bool,
|
134 |
max_new_tokens: int,
|
135 |
enable_thinking: bool,
|
136 |
+
enable_thinking_budget: bool, # NEWLY ADDED
|
137 |
+
thinking_budget: int, # NEWLY ADDED
|
138 |
):
|
139 |
"""
|
140 |
Runs a single turn of inference and yields the output stream for a gr.Chatbot.
|
|
|
143 |
prompt = chatbot[-1][0]
|
144 |
if (not image_input and not video_input and not prompt) or not prompt:
|
145 |
gr.Warning("A text prompt is required for generation.")
|
146 |
+
chatbot.pop(-1)
|
147 |
# MODIFICATION: Yield the current state and return to avoid errors
|
148 |
yield chatbot
|
149 |
return
|
150 |
|
|
|
|
|
|
|
|
|
151 |
content = []
|
152 |
if image_input:
|
153 |
content.append({"type": "image", "image": image_input})
|
|
|
157 |
content.append({"type": "video", "video": frames})
|
158 |
else:
|
159 |
gr.Warning("Failed to process the video file.")
|
160 |
+
chatbot.pop(-1)
|
161 |
yield chatbot
|
162 |
return
|
163 |
|
|
|
172 |
else:
|
173 |
input_ids, pixel_values, grid_thws = model.preprocess_inputs(messages=messages, add_generation_prompt=True, enable_thinking=enable_thinking)
|
174 |
except Exception as e:
|
175 |
+
gr.Warning(f"Error during input preprocessing: {e}")
|
176 |
+
chatbot.pop(-1)
|
177 |
yield chatbot
|
178 |
return
|
179 |
|
|
|
189 |
"eos_token_id": model.text_tokenizer.eos_token_id,
|
190 |
"pad_token_id": model.text_tokenizer.pad_token_id,
|
191 |
"streamer": streamer,
|
192 |
+
"use_cache": True,
|
193 |
+
"enable_thinking": enable_thinking,
|
194 |
+
"enable_thinking_budget": enable_thinking_budget,
|
195 |
+
"thinking_budget": thinking_budget
|
196 |
}
|
197 |
|
198 |
with torch.inference_mode():
|
|
|
219 |
chatbot[-1][1] = formatted_response
|
220 |
yield chatbot # Yield the final, formatted response
|
221 |
|
222 |
+
logger.info("\n[OVIS_CONV_START]")
|
223 |
+
[print(f'Q{i}:\n {request}\nA{i}:\n {answer}\n') for i, (request, answer) in enumerate(chatbot, 1)]
|
|
|
|
|
224 |
logger.info("[OVIS_CONV_END]")
|
225 |
|
226 |
|
|
|
|
|
|
|
227 |
# --- UI Helper Functions ---
|
228 |
def toggle_media_input(choice: str) -> Tuple:
|
229 |
"""Switches visibility between Image/Video inputs and their corresponding examples."""
|
|
|
234 |
|
235 |
|
236 |
# --- Build Gradio Application ---
|
|
|
237 |
def build_demo(model_path: str):
|
238 |
"""Builds the Gradio user interface for the model."""
|
239 |
global model, streamer
|
|
|
247 |
).to(device).eval()
|
248 |
|
249 |
text_tokenizer = model.text_tokenizer
|
250 |
+
streamer = MyTextIteratorStreamer(text_tokenizer, skip_prompt=True, skip_special_tokens=True)
|
251 |
|
252 |
print("Model loaded successfully.")
|
253 |
|
|
|
273 |
<center><font size=3><b>Ovis</b> has been open-sourced on <a href='https://huggingface.co/{model_path}'>😊 Huggingface</a> and <a href='https://github.com/AIDC-AI/Ovis'>🌟 GitHub</a>. If you find Ovis useful, a like❤️ or a star🌟 would be appreciated.</font></center>
|
274 |
"""
|
275 |
|
276 |
+
# --- START: Slider synchronization logic functions ---
|
277 |
+
def adjust_max_tokens(thinking_budget_val: int, max_new_tokens_val: int) -> gr.Slider:
|
278 |
+
"""Adjusts max_new_tokens to be at least thinking_budget + 128."""
|
279 |
+
new_max_tokens = max(max_new_tokens_val, thinking_budget_val + 128)
|
280 |
+
return gr.update(value=new_max_tokens)
|
281 |
+
|
282 |
+
def adjust_thinking_budget(max_new_tokens_val: int, thinking_budget_val: int) -> gr.Slider:
|
283 |
+
"""Adjusts thinking_budget to be at most max_new_tokens - 128."""
|
284 |
+
new_thinking_budget = min(thinking_budget_val, max_new_tokens_val - 128)
|
285 |
+
return gr.update(value=new_thinking_budget)
|
286 |
+
# --- END: Slider synchronization logic functions ---
|
287 |
+
|
288 |
prompt_input = gr.Textbox(label="Prompt", placeholder="Enter your text here and press ENTER", lines=1, container=False)
|
289 |
with gr.Blocks(theme=gr.themes.Ocean()) as demo:
|
290 |
gr.HTML(html_header)
|
291 |
+
gr.Markdown("Note: The Thinking Budget mechanism is enabled only when `Deep Thinking` and `Thinking Budget` are both checked. Could tune down `Thinking Budget` for faster generation in `Deep Thinking` mode.")
|
292 |
|
293 |
with gr.Row():
|
294 |
with gr.Column(scale=4):
|
|
|
298 |
|
299 |
with gr.Accordion("Generation Settings", open=True):
|
300 |
do_sample = gr.Checkbox(label="Enable Sampling (Do Sample)", value=True)
|
301 |
+
enable_thinking = gr.Checkbox(label="Enable Deep Thinking", value=True)
|
302 |
+
enable_thinking_budget = gr.Checkbox(label="Enable Thinking Budget", value=True)
|
303 |
+
max_new_tokens = gr.Slider(minimum=256, maximum=4096, value=2048, step=32, label="Max New Tokens")
|
304 |
+
thinking_budget = gr.Slider(minimum=128, maximum=3968, value=1024, step=32, label="Thinking Budget")
|
305 |
|
306 |
with gr.Column(visible=True) as image_examples_col:
|
307 |
gr.Examples(
|
|
|
325 |
generate_btn = gr.Button("Send", variant="primary")
|
326 |
clear_btn = gr.Button("Clear", variant="secondary")
|
327 |
|
328 |
+
# --- START: Event Handlers for UI Elements ---
|
329 |
+
|
330 |
input_type_radio.change(
|
331 |
fn=toggle_media_input,
|
332 |
inputs=input_type_radio,
|
333 |
outputs=[image_input, video_input, image_examples_col, video_examples_col]
|
334 |
)
|
335 |
|
336 |
+
# Event handlers for coupled sliders
|
337 |
+
thinking_budget.release(
|
338 |
+
fn=adjust_max_tokens,
|
339 |
+
inputs=[thinking_budget, max_new_tokens],
|
340 |
+
outputs=[max_new_tokens]
|
341 |
+
)
|
342 |
+
max_new_tokens.release(
|
343 |
+
fn=adjust_thinking_budget,
|
344 |
+
inputs=[max_new_tokens, thinking_budget],
|
345 |
+
outputs=[thinking_budget]
|
346 |
+
)
|
347 |
+
|
348 |
+
# MODIFICATION: Update run_inputs to include new controls
|
349 |
+
run_inputs = [chatbot, image_input, video_input, do_sample, max_new_tokens, enable_thinking, enable_thinking_budget, thinking_budget]
|
350 |
|
351 |
generat_click_event = generate_btn.click(submit_chat, [chatbot, prompt_input], [chatbot, prompt_input]).then(run_inference, run_inputs, chatbot)
|
352 |
submit_event = prompt_input.submit(submit_chat, [chatbot, prompt_input], [chatbot, prompt_input]).then(run_inference, run_inputs, chatbot)
|
353 |
|
354 |
+
# MODIFICATION: Update clear button to reset new controls
|
355 |
+
# clear_btn.click(
|
356 |
+
# fn=lambda: ([], None, None, "", "Image", True, 2048, True, True, 1024),
|
357 |
+
# outputs=[chatbot, image_input, video_input, prompt_input, input_type_radio, do_sample, max_new_tokens, enable_thinking, enable_thinking_budget, thinking_budget]
|
358 |
+
# ).then(
|
359 |
+
# fn=toggle_media_input,
|
360 |
+
# inputs=input_type_radio,
|
361 |
+
# outputs=[image_input, video_input, image_examples_col, video_examples_col]
|
362 |
+
# )
|
363 |
clear_btn.click(
|
364 |
+
fn=lambda: (list(), None, None, ""),
|
365 |
+
outputs=[chatbot, image_input, video_input, prompt_input]
|
|
|
|
|
|
|
|
|
366 |
)
|
367 |
+
# --- END: Event Handlers for UI Elements ---
|
368 |
|
369 |
return demo
|
370 |
|
371 |
+
|
372 |
# --- Main Execution Block ---
|
373 |
# def parse_args():
|
374 |
# parser = argparse.ArgumentParser(description="Gradio interface for a single Multimodal Large Language Model.")
|