Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -8,6 +8,91 @@ from transformers import AutoTokenizer, AutoModel, AutoConfig
|
|
8 |
import time
|
9 |
import re
|
10 |
from typing import List, Dict, Tuple, Optional
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
11 |
|
12 |
# Load model configuration to get special token IDs
|
13 |
config = AutoConfig.from_pretrained("Dream-org/Dream-v0-Instruct-7B", trust_remote_code=True)
|
@@ -27,25 +112,48 @@ print("Loading model...")
|
|
27 |
model = AutoModel.from_pretrained(
|
28 |
model_path,
|
29 |
torch_dtype=torch.bfloat16 if device == 'cuda' else torch.float32, # Use bfloat16 only on CUDA
|
30 |
-
trust_remote_code=True
|
|
|
31 |
)
|
32 |
model = model.to(device).eval()
|
33 |
print("Model loaded.")
|
34 |
|
35 |
# Constants from Dream's config/tokenizer
|
36 |
-
# Use attributes from loaded config/tokenizer objects
|
37 |
MASK_TOKEN = tokenizer.mask_token
|
38 |
-
MASK_ID =
|
39 |
-
PAD_ID =
|
40 |
-
EOS_ID =
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
41 |
# Make sure EOS_ID and PAD_ID are handled correctly; Dream uses the same ID for both
|
42 |
SPECIAL_TOKEN_IDS = {PAD_ID, EOS_ID, MASK_ID}
|
43 |
# Add other special tokens defined in tokenizer_config.json if needed for hiding
|
44 |
# Get IDs for im_start, im_end etc. if they should also be hidden/handled specially
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
SPECIAL_TOKEN_IDS.add(
|
|
|
|
|
|
|
|
|
|
|
|
|
49 |
|
50 |
# --- Helper Functions ---
|
51 |
|
@@ -61,25 +169,57 @@ def parse_constraints(constraints_text: str) -> Dict[int, List[int]]:
|
|
61 |
|
62 |
parts = constraints_text.split(',')
|
63 |
for part in parts:
|
|
|
64 |
if ':' not in part:
|
65 |
continue
|
66 |
pos_str, word = part.split(':', 1)
|
67 |
try:
|
68 |
# Position relative to the start of the *generation*
|
69 |
pos = int(pos_str.strip())
|
70 |
-
word = word.strip()
|
71 |
-
# Tokenize the word -
|
72 |
-
#
|
73 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
74 |
|
75 |
if token_ids and pos >= 0:
|
76 |
constraints[pos] = token_ids
|
|
|
|
|
77 |
except ValueError:
|
|
|
78 |
continue # Ignore malformed constraint parts
|
79 |
except Exception as e:
|
80 |
print(f"Warning: Error processing constraint '{part}': {e}")
|
81 |
continue
|
82 |
|
|
|
83 |
return constraints
|
84 |
|
85 |
|
@@ -95,23 +235,45 @@ def format_chat_history(history: List[List[Optional[str]]]) -> List[Dict[str, st
|
|
95 |
Formatted list of message dictionaries for tokenizer.apply_chat_template.
|
96 |
"""
|
97 |
messages = []
|
98 |
-
# Check if the first message is a system prompt, handle accordingly if needed
|
99 |
-
# Based on Dream's examples, the template adds a default system prompt if none exists.
|
100 |
-
# If history starts with System, it should be handled by the template.
|
101 |
-
# Let's assume the template handles the system prompt correctly.
|
102 |
-
|
103 |
for user_msg, assistant_msg in history:
|
104 |
if user_msg: # Defensive check
|
105 |
messages.append({"role": "user", "content": user_msg})
|
106 |
# Add assistant message only if it exists (it won't for the last turn before generation)
|
107 |
if assistant_msg:
|
108 |
messages.append({"role": "assistant", "content": assistant_msg})
|
109 |
-
|
110 |
return messages
|
111 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
112 |
# --- Core Generation Logic with Live Visualization ---
|
113 |
|
114 |
@spaces.GPU # Decorator for Hugging Face Spaces GPU usage
|
|
|
115 |
def generate_dream_response(
|
116 |
history: List[List[Optional[str]]],
|
117 |
gen_length: int,
|
@@ -125,7 +287,7 @@ def generate_dream_response(
|
|
125 |
visualization_delay: float
|
126 |
) -> List[Tuple[str, str]]:
|
127 |
"""
|
128 |
-
Generates text using the Dream model and yields visualization states live.
|
129 |
|
130 |
Args:
|
131 |
history: Chat history.
|
@@ -133,21 +295,20 @@ def generate_dream_response(
|
|
133 |
steps: Number of diffusion steps.
|
134 |
constraints_text: User-provided constraints string.
|
135 |
temperature: Sampling temperature.
|
136 |
-
top_p: Top-p sampling nucleus.
|
137 |
-
top_k: Top-k sampling.
|
138 |
alg: Remasking algorithm ('origin', 'maskgit_plus', 'topk_margin', 'entropy').
|
139 |
alg_temp: Temperature for confidence-based algorithms.
|
140 |
visualization_delay: Delay between visualization steps.
|
141 |
|
142 |
Yields:
|
143 |
Tuple[List[List[Optional[str]]], List[Tuple[str, Optional[str]]], str]:
|
144 |
-
- Updated history
|
145 |
-
- Visualization data for HighlightedText
|
146 |
-
- Final response text (
|
147 |
"""
|
148 |
|
149 |
if not history or not history[-1][0]:
|
150 |
-
# No user message to respond to
|
151 |
yield history, [("No input message found.", "red")], ""
|
152 |
return
|
153 |
|
@@ -167,90 +328,275 @@ def generate_dream_response(
|
|
167 |
add_generation_prompt=True # Important for instruct models
|
168 |
)
|
169 |
input_ids = inputs.input_ids.to(device)
|
170 |
-
|
171 |
prompt_length = input_ids.shape[1]
|
172 |
except Exception as e:
|
173 |
print(f"Error applying chat template: {e}")
|
174 |
yield history, [("Error preparing input.", "red")], ""
|
175 |
return
|
176 |
|
177 |
-
#
|
178 |
-
|
179 |
-
#
|
180 |
-
|
181 |
-
|
|
|
182 |
|
183 |
-
# --- 2.
|
184 |
-
|
185 |
-
step_sequence_history: List[torch.Tensor] = []
|
186 |
-
previous_step_tokens = None # Keep track of the previous step's state
|
187 |
|
188 |
-
#
|
189 |
-
|
190 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
191 |
|
192 |
-
|
193 |
-
|
194 |
-
|
195 |
-
|
196 |
-
|
197 |
-
current_x = x.clone() # Work on a copy
|
198 |
|
199 |
-
|
200 |
-
|
201 |
-
abs_end_pos = abs_start_pos + len(word_token_ids)
|
202 |
|
203 |
-
|
204 |
-
|
205 |
-
|
206 |
-
|
207 |
-
|
208 |
-
|
209 |
-
|
210 |
-
|
211 |
-
|
212 |
-
|
213 |
-
|
214 |
-
|
215 |
-
|
216 |
-
|
217 |
-
|
218 |
-
|
219 |
-
|
220 |
-
|
221 |
-
|
222 |
-
|
223 |
-
|
224 |
-
|
225 |
-
|
226 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
227 |
|
228 |
-
# Initial masked state for visualization
|
229 |
-
initial_generated_state = torch.full((gen_length,), MASK_ID, dtype=torch.long)
|
230 |
-
# Apply constraints to the *initial* visual state if they start at pos 0
|
231 |
-
temp_initial_x = torch.cat((input_ids[0], initial_generated_state.to(device)), dim=0).unsqueeze(0)
|
232 |
-
initial_vis_x = live_visualization_hook(None, temp_initial_x, None) # Apply constraints via hook logic
|
233 |
-
step_sequence_history.insert(0, initial_vis_x[0, prompt_length:].cpu()) # Prepend initial state
|
234 |
-
|
235 |
-
output = model.diffusion_generate(
|
236 |
-
input_ids,
|
237 |
-
attention_mask=attention_mask,
|
238 |
-
max_new_tokens=gen_length,
|
239 |
-
output_history=False, # We capture history via the hook
|
240 |
-
return_dict_in_generate=True,
|
241 |
-
steps=steps,
|
242 |
-
temperature=temperature,
|
243 |
-
top_p=top_p if top_p is not None and top_p < 1.0 else None, # Ensure top_p < 1 or None
|
244 |
-
top_k=top_k if top_k is not None and top_k > 0 else None, # Ensure top_k > 0 or None
|
245 |
-
alg=alg,
|
246 |
-
alg_temp=alg_temp if alg in ['maskgit_plus', 'topk_margin', 'entropy'] else None, # Only relevant for some algs
|
247 |
-
generation_tokens_hook_func=live_visualization_hook
|
248 |
-
)
|
249 |
end_time = time.time()
|
250 |
print(f"Dream generation finished in {end_time - start_time:.2f} seconds.")
|
251 |
|
252 |
-
# ---
|
253 |
-
final_sequence =
|
254 |
response_tokens = final_sequence[prompt_length:]
|
255 |
|
256 |
# Decode the final response text
|
@@ -260,83 +606,57 @@ def generate_dream_response(
|
|
260 |
clean_up_tokenization_spaces=True
|
261 |
).strip()
|
262 |
|
263 |
-
# Update history with the final response
|
264 |
-
|
265 |
|
266 |
-
|
267 |
-
|
268 |
-
|
269 |
-
|
270 |
-
yield history, [("Error during generation.", "red")], ""
|
271 |
-
return
|
272 |
-
|
273 |
-
# --- 5. Stream Visualization ---
|
274 |
-
print(f"Streaming {len(step_sequence_history)} visualization steps...")
|
275 |
-
previous_tokens_vis = None
|
276 |
-
for i, current_tokens_vis in enumerate(step_sequence_history):
|
277 |
-
# print(f" Step {i}: {current_tokens_vis.tolist()}") # Debug
|
278 |
-
vis_data = []
|
279 |
-
current_decoded_tokens = []
|
280 |
-
|
281 |
-
# Compare current step tokens with previous step tokens
|
282 |
for j in range(gen_length):
|
283 |
-
current_tok_id =
|
284 |
previous_tok_id = previous_tokens_vis[j].item() if previous_tokens_vis is not None else MASK_ID
|
285 |
|
286 |
-
# Decode token - handle potential errors for single IDs if needed
|
287 |
try:
|
288 |
-
# Use skip_special_tokens=False here to see the actual tokens
|
289 |
decoded_token = tokenizer.decode([current_tok_id], skip_special_tokens=False)
|
290 |
-
|
291 |
-
if current_tok_id == MASK_ID:
|
292 |
-
display_token = MASK_TOKEN
|
293 |
-
else:
|
294 |
-
display_token = decoded_token
|
295 |
-
|
296 |
except Exception:
|
297 |
display_token = f"[ID:{current_tok_id}]" # Fallback
|
298 |
|
299 |
-
# Determine color and handle hiding of special tokens (like LLaDA demo)
|
300 |
color = None
|
301 |
token_to_display = display_token
|
302 |
|
303 |
if current_tok_id == MASK_ID:
|
304 |
-
color = "#444444"
|
305 |
-
elif previous_tok_id == MASK_ID:
|
306 |
-
|
307 |
-
|
308 |
-
|
309 |
-
|
310 |
-
|
311 |
-
|
312 |
-
if
|
313 |
-
# Hide by making it empty or using a background color - empty string is simpler
|
314 |
token_to_display = ""
|
315 |
-
color =
|
316 |
|
317 |
-
|
318 |
-
|
319 |
-
vis_data.append((token_to_display, color))
|
320 |
-
elif len(vis_data) > 0 and isinstance(vis_data[-1], tuple):
|
321 |
-
# If hidden, and previous was text, add a space for visual separation?
|
322 |
-
# This might complicate things, let's omit for now.
|
323 |
-
pass
|
324 |
-
# elif len(vis_data) == 0: # If first token is hidden
|
325 |
-
# vis_data.append(("", None)) # Placeholder?
|
326 |
|
327 |
-
#
|
328 |
-
|
|
|
329 |
|
330 |
-
# Yield the current visualization state
|
331 |
-
yield history, vis_data, final_response_text
|
332 |
|
333 |
-
|
334 |
-
|
335 |
-
|
336 |
-
|
|
|
|
|
|
|
337 |
|
338 |
|
339 |
-
# --- Gradio UI ---
|
340 |
css = '''
|
341 |
.category-legend{display:none}
|
342 |
button{min-height: 60px}
|
@@ -346,11 +666,13 @@ def create_chatbot_demo():
|
|
346 |
gr.Markdown("# Dream 7B - Diffusion Language Model Demo")
|
347 |
gr.Markdown(
|
348 |
"[[Model Card](https://huggingface.co/Dream-org/Dream-v0-Instruct-7B)] "
|
349 |
-
"[[Blog](https://hkunlp.github.io/blog/2025/dream/)]"
|
350 |
)
|
351 |
|
352 |
# STATE MANAGEMENT
|
353 |
-
chat_history = gr.State([])
|
|
|
|
|
354 |
|
355 |
# UI COMPONENTS
|
356 |
with gr.Row():
|
@@ -359,7 +681,8 @@ def create_chatbot_demo():
|
|
359 |
label="Conversation",
|
360 |
height=500,
|
361 |
show_copy_button=True,
|
362 |
-
bubble_full_width=False
|
|
|
363 |
)
|
364 |
|
365 |
# Message input
|
@@ -386,9 +709,17 @@ def create_chatbot_demo():
|
|
386 |
output_vis = gr.HighlightedText(
|
387 |
label="Denoising Process Visualization",
|
388 |
combine_adjacent=False,
|
389 |
-
show_legend=True,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
390 |
)
|
391 |
|
|
|
392 |
# Advanced generation settings
|
393 |
with gr.Accordion("Generation Settings", open=False):
|
394 |
with gr.Row():
|
@@ -403,21 +734,21 @@ def create_chatbot_demo():
|
|
403 |
with gr.Row():
|
404 |
temperature = gr.Slider(
|
405 |
minimum=0.0, maximum=1.0, value=0.4, step=0.05,
|
406 |
-
label="Temperature"
|
407 |
)
|
408 |
alg_temp = gr.Slider(
|
409 |
minimum=0.0, maximum=1.0, value=0.1, step=0.05,
|
410 |
-
label="Remasking Temp (
|
411 |
)
|
412 |
|
413 |
with gr.Row():
|
414 |
top_p = gr.Slider(
|
415 |
minimum=0.0, maximum=1.0, value=0.95, step=0.05,
|
416 |
-
label="Top-P (0
|
417 |
)
|
418 |
top_k = gr.Slider(
|
419 |
minimum=0, maximum=200, value=0, step=5,
|
420 |
-
label="Top-K (0
|
421 |
)
|
422 |
|
423 |
with gr.Row():
|
@@ -429,76 +760,77 @@ def create_chatbot_demo():
|
|
429 |
|
430 |
with gr.Row():
|
431 |
visualization_delay = gr.Slider(
|
432 |
-
minimum=0.0, maximum=0.5, value=0.
|
433 |
label="Visualization Delay (seconds)"
|
434 |
)
|
435 |
|
436 |
# Clear button
|
437 |
clear_btn = gr.Button("Clear Conversation")
|
438 |
|
439 |
-
# Current response text box (hidden, maybe useful for debugging)
|
440 |
-
# current_response = gr.Textbox(visible=False)
|
441 |
-
|
442 |
# --- Event Handlers ---
|
443 |
|
444 |
-
def add_user_message_to_history(message: str,
|
445 |
"""Adds user message, clears input, prepares for bot response."""
|
446 |
if not message.strip():
|
447 |
gr.Warning("Please enter a message.")
|
448 |
-
|
|
|
449 |
|
450 |
# Add user message with placeholder for bot response
|
451 |
-
|
452 |
-
# Return updated history for chatbot, empty input
|
453 |
-
return
|
454 |
-
|
455 |
|
456 |
def clear_conversation():
|
457 |
-
"""Clears the chat history and
|
458 |
-
return [], [], "", []
|
459 |
|
460 |
# --- Connect UI elements ---
|
461 |
|
462 |
# Define the inputs for the generation function once
|
463 |
generation_inputs = [
|
464 |
-
|
465 |
temperature, top_p, top_k, remasking_strategy, alg_temp,
|
466 |
visualization_delay
|
467 |
]
|
468 |
# Define the outputs for the generation function
|
469 |
-
|
|
|
|
|
470 |
|
471 |
# Handle Textbox Submission (Enter key)
|
472 |
submit_listener = user_input.submit(
|
473 |
fn=add_user_message_to_history,
|
474 |
-
inputs=[user_input,
|
475 |
-
outputs=[
|
476 |
)
|
477 |
# Chain the bot response generation after the user message is added
|
478 |
submit_listener.then(
|
479 |
fn=generate_dream_response,
|
480 |
inputs=generation_inputs,
|
481 |
-
outputs=generation_outputs # Step 2: Generate response and stream vis
|
|
|
482 |
)
|
483 |
|
484 |
# Handle Send Button Click
|
485 |
click_listener = send_btn.click(
|
486 |
fn=add_user_message_to_history,
|
487 |
-
inputs=[user_input,
|
488 |
-
outputs=[
|
489 |
)
|
490 |
# Chain the bot response generation after the user message is added
|
491 |
click_listener.then(
|
492 |
fn=generate_dream_response,
|
493 |
inputs=generation_inputs,
|
494 |
-
outputs=generation_outputs # Step 2: Generate response and stream vis
|
|
|
495 |
)
|
496 |
|
497 |
-
# Clear Button Action
|
498 |
clear_btn.click(
|
499 |
clear_conversation,
|
500 |
inputs=[],
|
501 |
-
outputs=[
|
502 |
)
|
503 |
|
504 |
return demo
|
@@ -507,4 +839,4 @@ def create_chatbot_demo():
|
|
507 |
if __name__ == "__main__":
|
508 |
demo = create_chatbot_demo()
|
509 |
# Use queue for handling multiple users and streaming
|
510 |
-
demo.queue().launch(debug=True, share=
|
|
|
8 |
import time
|
9 |
import re
|
10 |
from typing import List, Dict, Tuple, Optional
|
11 |
+
import torch.distributions as dists # Added import
|
12 |
+
|
13 |
+
# --- START: Copied Helper functions from generation_utils.py ---
|
14 |
+
# These are needed because we are reimplementing the sampling loop locally.
|
15 |
+
|
16 |
+
def top_p_logits(logits, top_p=None):
|
17 |
+
""" Applies top-p filtering to logits. """
|
18 |
+
if top_p is None or top_p >= 1.0:
|
19 |
+
return logits
|
20 |
+
sorted_logits, sorted_indices = torch.sort(logits, descending=True)
|
21 |
+
cumulative_probs = torch.cumsum(F.softmax(sorted_logits, dim=-1), dim=-1)
|
22 |
+
sorted_indices_to_remove = cumulative_probs > top_p
|
23 |
+
# Shift the indices to the right to keep the first token above the threshold
|
24 |
+
sorted_indices_to_remove[..., 1:] = sorted_indices_to_remove[..., :-1].clone()
|
25 |
+
sorted_indices_to_remove[..., 0] = 0
|
26 |
+
|
27 |
+
mask = torch.zeros_like(logits, dtype=torch.bool, device=logits.device)
|
28 |
+
mask = mask.scatter_(-1, sorted_indices, sorted_indices_to_remove)
|
29 |
+
logits = logits.masked_fill(mask, torch.finfo(logits.dtype).min)
|
30 |
+
return logits
|
31 |
+
|
32 |
+
def top_k_logits(logits, top_k=None):
|
33 |
+
""" Applies top-k filtering to logits. """
|
34 |
+
if top_k is None or top_k <= 0:
|
35 |
+
return logits
|
36 |
+
top_k = min(top_k, logits.size(-1)) # Safety check
|
37 |
+
# Remove all tokens with a probability less than the last token of the top-k
|
38 |
+
indices_to_remove = logits < torch.topk(logits, top_k)[0][..., -1, None]
|
39 |
+
logits = logits.masked_fill(indices_to_remove, torch.finfo(logits.dtype).min)
|
40 |
+
return logits
|
41 |
+
|
42 |
+
def sample_tokens(logits, temperature=0.0, top_p=None, top_k=None, margin_confidence=False, neg_entropy=False):
|
43 |
+
""" Samples tokens based on logits and calculates confidence. """
|
44 |
+
if temperature > 0:
|
45 |
+
logits = logits / temperature
|
46 |
+
if top_p is not None and top_p < 1.0: # Apply top_p if valid
|
47 |
+
logits = top_p_logits(logits, top_p)
|
48 |
+
if top_k is not None and top_k > 0: # Apply top_k if valid
|
49 |
+
logits = top_k_logits(logits, top_k)
|
50 |
+
|
51 |
+
# Ensure logits are not all -inf after filtering, if so, sample uniformly? Or handle error.
|
52 |
+
# For simplicity, assume valid logits after filtering. If not, sampling might fail.
|
53 |
+
# Add a small epsilon to prevent log(0) or issues with all -inf logits
|
54 |
+
logits = torch.where(logits == torch.finfo(logits.dtype).min, torch.full_like(logits, -1e9), logits)
|
55 |
+
|
56 |
+
|
57 |
+
probs = torch.softmax(logits, dim=-1)
|
58 |
+
|
59 |
+
if temperature > 0:
|
60 |
+
try:
|
61 |
+
# Check for NaNs or Infs in probs before sampling
|
62 |
+
if torch.isnan(probs).any() or torch.isinf(probs).any():
|
63 |
+
print("Warning: NaN or Inf detected in probabilities before sampling. Attempting to recover.")
|
64 |
+
# Simple recovery: Sample from uniform distribution or highest prob token
|
65 |
+
probs = torch.nan_to_num(probs, nan=0.0, posinf=0.0, neginf=0.0)
|
66 |
+
if probs.sum() == 0: # If all probabilities became zero
|
67 |
+
print("Warning: All probabilities became zero. Sampling uniformly.")
|
68 |
+
probs = torch.ones_like(probs) / probs.shape[-1]
|
69 |
+
else:
|
70 |
+
probs = probs / probs.sum(dim=-1, keepdim=True) # Re-normalize
|
71 |
+
|
72 |
+
x0 = dists.Categorical(probs=probs).sample()
|
73 |
+
confidence = torch.gather(probs, -1, x0.unsqueeze(-1)).squeeze(-1)
|
74 |
+
except Exception as e: # Catch broader exceptions during sampling
|
75 |
+
print(f"Warning: Error during Categorical sampling: {e}. Falling back to argmax.")
|
76 |
+
confidence, x0 = probs.max(dim=-1)
|
77 |
+
else:
|
78 |
+
confidence, x0 = probs.max(dim=-1)
|
79 |
+
|
80 |
+
if margin_confidence:
|
81 |
+
sorted_probs, _ = torch.sort(probs, dim=-1, descending=True)
|
82 |
+
# Ensure there are at least 2 probabilities to compare
|
83 |
+
top1_probs = sorted_probs[..., 0]
|
84 |
+
top2_probs = sorted_probs[..., 1] if sorted_probs.shape[-1] > 1 else top1_probs # Handle case with only 1 possible token
|
85 |
+
confidence = top1_probs - top2_probs
|
86 |
+
|
87 |
+
if neg_entropy:
|
88 |
+
epsilon = 1e-10
|
89 |
+
log_probs = torch.log(probs + epsilon)
|
90 |
+
confidence = torch.sum(probs * log_probs, dim=-1) # Should be negative entropy
|
91 |
+
|
92 |
+
return confidence, x0
|
93 |
+
|
94 |
+
# --- END: Copied Helper functions ---
|
95 |
+
|
96 |
|
97 |
# Load model configuration to get special token IDs
|
98 |
config = AutoConfig.from_pretrained("Dream-org/Dream-v0-Instruct-7B", trust_remote_code=True)
|
|
|
112 |
model = AutoModel.from_pretrained(
|
113 |
model_path,
|
114 |
torch_dtype=torch.bfloat16 if device == 'cuda' else torch.float32, # Use bfloat16 only on CUDA
|
115 |
+
trust_remote_code=True,
|
116 |
+
# attn_implementation="flash_attention_2" # Optional: Speed up if FA2 is available
|
117 |
)
|
118 |
model = model.to(device).eval()
|
119 |
print("Model loaded.")
|
120 |
|
121 |
# Constants from Dream's config/tokenizer
|
|
|
122 |
MASK_TOKEN = tokenizer.mask_token
|
123 |
+
MASK_ID = tokenizer.mask_token_id # Use tokenizer's mask_token_id directly
|
124 |
+
PAD_ID = tokenizer.pad_token_id # Use tokenizer's pad_token_id
|
125 |
+
EOS_ID = tokenizer.eos_token_id # Use tokenizer's eos_token_id
|
126 |
+
# Use attributes from loaded config/tokenizer objects
|
127 |
+
# MASK_ID = config.mask_token_id # Can use this too, should be consistent
|
128 |
+
# PAD_ID = config.pad_token_id
|
129 |
+
# EOS_ID = config.eos_token_id
|
130 |
+
|
131 |
+
# Ensure mask_token_id is correctly identified
|
132 |
+
if MASK_ID is None:
|
133 |
+
print("Warning: Mask token ID not found in config/tokenizer. Trying to fetch from tokenizer...")
|
134 |
+
# Try getting from tokenizer directly if config doesn't have it or it's None
|
135 |
+
mask_token_special = tokenizer.mask_token
|
136 |
+
if mask_token_special:
|
137 |
+
MASK_ID = tokenizer.convert_tokens_to_ids(mask_token_special)
|
138 |
+
print(f"Found MASK_ID from tokenizer: {MASK_ID}")
|
139 |
+
else:
|
140 |
+
# Fallback or raise error if still not found
|
141 |
+
raise ValueError("Cannot determine MASK_ID. Check model's tokenizer configuration.")
|
142 |
+
|
143 |
# Make sure EOS_ID and PAD_ID are handled correctly; Dream uses the same ID for both
|
144 |
SPECIAL_TOKEN_IDS = {PAD_ID, EOS_ID, MASK_ID}
|
145 |
# Add other special tokens defined in tokenizer_config.json if needed for hiding
|
146 |
# Get IDs for im_start, im_end etc. if they should also be hidden/handled specially
|
147 |
+
try:
|
148 |
+
IM_START_ID = tokenizer.convert_tokens_to_ids("<|im_start|>")
|
149 |
+
IM_END_ID = tokenizer.convert_tokens_to_ids("<|im_end|>")
|
150 |
+
SPECIAL_TOKEN_IDS.add(IM_START_ID)
|
151 |
+
SPECIAL_TOKEN_IDS.add(IM_END_ID)
|
152 |
+
except KeyError:
|
153 |
+
print("Warning: <|im_start|> or <|im_end|> not found in tokenizer vocab.")
|
154 |
+
IM_START_ID = None
|
155 |
+
IM_END_ID = None
|
156 |
+
|
157 |
|
158 |
# --- Helper Functions ---
|
159 |
|
|
|
169 |
|
170 |
parts = constraints_text.split(',')
|
171 |
for part in parts:
|
172 |
+
part = part.strip() # Remove leading/trailing whitespace from the part itself
|
173 |
if ':' not in part:
|
174 |
continue
|
175 |
pos_str, word = part.split(':', 1)
|
176 |
try:
|
177 |
# Position relative to the start of the *generation*
|
178 |
pos = int(pos_str.strip())
|
179 |
+
word = word.strip() # Strip whitespace from word
|
180 |
+
# Tokenize the word - Dream tokenizer handles spaces well typically.
|
181 |
+
# Let's check if the word starts with a space implicitly or needs one.
|
182 |
+
# Standard tokenizers often need a space prefix if not at the start.
|
183 |
+
# Test: tokenizer.encode(" world") vs tokenizer.encode("world")
|
184 |
+
# Dream often encodes ' world' differently from 'world'.
|
185 |
+
# Assume we want the word as it would appear mid-sentence unless pos is 0.
|
186 |
+
token_ids = tokenizer.encode(word, add_special_tokens=False)
|
187 |
+
# Add space prefix if needed based on position? This is tricky.
|
188 |
+
# Let's assume the user provides the word how they want it tokenized,
|
189 |
+
# potentially including a leading space if necessary.
|
190 |
+
# Example: " 5: word" might be tokenized differently than "5:word".
|
191 |
+
# Simplest approach: Tokenize exactly what the user provided.
|
192 |
+
# Let's refine: add space prefix automatically if pos > 0, unless word already starts with space?
|
193 |
+
# This seems more robust for typical usage.
|
194 |
+
if pos > 0 and not word.startswith(" "):
|
195 |
+
token_ids_with_space = tokenizer.encode(" " + word, add_special_tokens=False)
|
196 |
+
# Check if adding space actually changes tokenization significantly
|
197 |
+
# Heuristic: if the first token ID changes, use the space-prefixed version.
|
198 |
+
first_token_no_space = tokenizer.encode(word, add_special_tokens=False)[0] if token_ids else None
|
199 |
+
first_token_with_space = tokenizer.encode(" " + word, add_special_tokens=False)[0] if token_ids_with_space else None
|
200 |
+
|
201 |
+
if first_token_no_space != first_token_with_space:
|
202 |
+
token_ids = token_ids_with_space
|
203 |
+
# If tokenization doesn't change much, maybe stick to original? Less surprising.
|
204 |
+
# Let's stick to adding the space if pos > 0 for consistency, like original code.
|
205 |
+
token_ids = tokenizer.encode(" " + word, add_special_tokens=False)
|
206 |
+
|
207 |
+
elif pos == 0:
|
208 |
+
token_ids = tokenizer.encode(word, add_special_tokens=False)
|
209 |
+
|
210 |
|
211 |
if token_ids and pos >= 0:
|
212 |
constraints[pos] = token_ids
|
213 |
+
elif not token_ids:
|
214 |
+
print(f"Warning: Could not tokenize constraint word '{word}'")
|
215 |
except ValueError:
|
216 |
+
print(f"Warning: Invalid position '{pos_str}' in constraint part '{part}'")
|
217 |
continue # Ignore malformed constraint parts
|
218 |
except Exception as e:
|
219 |
print(f"Warning: Error processing constraint '{part}': {e}")
|
220 |
continue
|
221 |
|
222 |
+
print(f"Parsed constraints: {constraints}") # Debugging
|
223 |
return constraints
|
224 |
|
225 |
|
|
|
235 |
Formatted list of message dictionaries for tokenizer.apply_chat_template.
|
236 |
"""
|
237 |
messages = []
|
|
|
|
|
|
|
|
|
|
|
238 |
for user_msg, assistant_msg in history:
|
239 |
if user_msg: # Defensive check
|
240 |
messages.append({"role": "user", "content": user_msg})
|
241 |
# Add assistant message only if it exists (it won't for the last turn before generation)
|
242 |
if assistant_msg:
|
243 |
messages.append({"role": "assistant", "content": assistant_msg})
|
|
|
244 |
return messages
|
245 |
|
246 |
+
def apply_constraints_to_state(
|
247 |
+
x: torch.Tensor,
|
248 |
+
prompt_length: int,
|
249 |
+
total_length: int,
|
250 |
+
parsed_constraints: Dict[int, List[int]],
|
251 |
+
current_step: Optional[int] = None # For logging/debugging
|
252 |
+
) -> torch.Tensor:
|
253 |
+
"""Applies constraints directly to the state tensor `x`."""
|
254 |
+
modified_x = x.clone() # Work on a copy to avoid modifying original if needed elsewhere
|
255 |
+
for rel_pos, word_token_ids in parsed_constraints.items():
|
256 |
+
abs_start_pos = prompt_length + rel_pos
|
257 |
+
abs_end_pos = abs_start_pos + len(word_token_ids)
|
258 |
+
|
259 |
+
# Ensure the constraint fits within the generation length
|
260 |
+
if abs_start_pos < total_length and abs_end_pos <= total_length:
|
261 |
+
try:
|
262 |
+
constraint_tensor = torch.tensor(word_token_ids, dtype=torch.long, device=modified_x.device)
|
263 |
+
# Force the constraint tokens onto the sequence
|
264 |
+
modified_x[0, abs_start_pos:abs_end_pos] = constraint_tensor
|
265 |
+
# print(f"Debug (Step {current_step}): Applied constraint {tokenizer.decode(word_token_ids)} at pos {rel_pos}") # Debug
|
266 |
+
except IndexError:
|
267 |
+
print(f"Warning (Step {current_step}): Constraint at {rel_pos} ('{tokenizer.decode(word_token_ids)}') goes out of bounds.")
|
268 |
+
except Exception as e:
|
269 |
+
print(f"Warning (Step {current_step}): Failed to apply constraint at {rel_pos}: {e}")
|
270 |
+
return modified_x
|
271 |
+
|
272 |
+
|
273 |
# --- Core Generation Logic with Live Visualization ---
|
274 |
|
275 |
@spaces.GPU # Decorator for Hugging Face Spaces GPU usage
|
276 |
+
@torch.no_grad() # Ensure no gradients are computed during generation
|
277 |
def generate_dream_response(
|
278 |
history: List[List[Optional[str]]],
|
279 |
gen_length: int,
|
|
|
287 |
visualization_delay: float
|
288 |
) -> List[Tuple[str, str]]:
|
289 |
"""
|
290 |
+
Generates text using the Dream model step-by-step and yields visualization states live.
|
291 |
|
292 |
Args:
|
293 |
history: Chat history.
|
|
|
295 |
steps: Number of diffusion steps.
|
296 |
constraints_text: User-provided constraints string.
|
297 |
temperature: Sampling temperature.
|
298 |
+
top_p: Top-p sampling nucleus. Clamp to < 1.0 or None.
|
299 |
+
top_k: Top-k sampling. Clamp to > 0 or None.
|
300 |
alg: Remasking algorithm ('origin', 'maskgit_plus', 'topk_margin', 'entropy').
|
301 |
alg_temp: Temperature for confidence-based algorithms.
|
302 |
visualization_delay: Delay between visualization steps.
|
303 |
|
304 |
Yields:
|
305 |
Tuple[List[List[Optional[str]]], List[Tuple[str, Optional[str]]], str]:
|
306 |
+
- Updated history (may be intermediate until final response)
|
307 |
+
- Visualization data for HighlightedText for the current step
|
308 |
+
- Intermediate or Final response text (yielded repeatedly)
|
309 |
"""
|
310 |
|
311 |
if not history or not history[-1][0]:
|
|
|
312 |
yield history, [("No input message found.", "red")], ""
|
313 |
return
|
314 |
|
|
|
328 |
add_generation_prompt=True # Important for instruct models
|
329 |
)
|
330 |
input_ids = inputs.input_ids.to(device)
|
331 |
+
prompt_attention_mask = inputs.attention_mask.to(device) # Mask for the prompt part
|
332 |
prompt_length = input_ids.shape[1]
|
333 |
except Exception as e:
|
334 |
print(f"Error applying chat template: {e}")
|
335 |
yield history, [("Error preparing input.", "red")], ""
|
336 |
return
|
337 |
|
338 |
+
# --- Config parameters for the loop ---
|
339 |
+
eps = 1e-3 # Default from DreamGenerationConfig, make configurable if needed
|
340 |
+
# Ensure top_p and top_k have valid values for filtering functions
|
341 |
+
top_p_val = top_p if top_p is not None and top_p < 1.0 else None
|
342 |
+
top_k_val = top_k if top_k is not None and top_k > 0 else None
|
343 |
+
alg_temp_val = alg_temp if alg in ['maskgit_plus', 'topk_margin', 'entropy'] else None
|
344 |
|
345 |
+
# --- 2. Initialize Generation State ---
|
346 |
+
total_length = prompt_length + gen_length
|
|
|
|
|
347 |
|
348 |
+
# Initial state: prompt + MASK tokens
|
349 |
+
initial_generation_part = torch.full((1, gen_length), MASK_ID, dtype=torch.long, device=device)
|
350 |
+
x = torch.cat((input_ids, initial_generation_part), dim=1)
|
351 |
+
|
352 |
+
# Prepare full attention mask (assuming full attention over generated part initially)
|
353 |
+
generation_attention_mask = torch.ones((1, gen_length), dtype=torch.long, device=device)
|
354 |
+
full_attention_mask = torch.cat((prompt_attention_mask, generation_attention_mask), dim=1)
|
355 |
+
|
356 |
+
# Check if model needs specific attention mask format (e.g., causal for prompt?)
|
357 |
+
# The original `diffusion_generate` handles this internally. Replicating requires care.
|
358 |
+
# Based on `_sample`, it prepares a broadcastable mask if padding exists, else uses "full".
|
359 |
+
# Let's assume "full" attention is okay for Dream's purpose here, as mask tokens don't depend on future masks.
|
360 |
+
# If the base model *requires* causal masking internally even with diffusion, this might need adjustment.
|
361 |
+
# For simplicity, using a full mask (ones) over the whole sequence.
|
362 |
+
# The model's internal attention should handle causality if needed.
|
363 |
+
# Let's stick to the simpler full mask preparation from the original code when no padding.
|
364 |
+
if torch.any(full_attention_mask == 0): # Handle padding if present (shouldn't be with template?)
|
365 |
+
tok_idx = full_attention_mask.long().cumsum(-1) - 1
|
366 |
+
tok_idx.masked_fill_(full_attention_mask == 0, 0) # Use 0 for padding index? Or 1? Check original. Original used 1.
|
367 |
+
tok_idx.masked_fill_(full_attention_mask == 0, 1)
|
368 |
+
attention_mask_for_model = torch.logical_and(
|
369 |
+
full_attention_mask.unsqueeze(1).unsqueeze(-2),
|
370 |
+
full_attention_mask.unsqueeze(1).unsqueeze(-1),
|
371 |
+
) # Shape [B, 1, N, N]
|
372 |
+
else:
|
373 |
+
tok_idx = None
|
374 |
+
attention_mask_for_model = None # Let the model handle full attention if mask is None
|
375 |
+
|
376 |
+
# Timesteps for diffusion
|
377 |
+
timesteps = torch.linspace(1, eps, steps + 1, device=device)
|
378 |
+
|
379 |
+
# Apply initial constraints (before first step)
|
380 |
+
x = apply_constraints_to_state(x, prompt_length, total_length, parsed_constraints, current_step=-1) # Step -1 for initial
|
381 |
+
|
382 |
+
# --- 3. Visualization Setup ---
|
383 |
+
previous_tokens_vis = None # Keep track of the previous step's state for coloring
|
384 |
+
final_response_text = "" # Store the final decoded text
|
385 |
+
history_copy = [list(item) for item in history] # Make a mutable copy
|
386 |
+
|
387 |
+
# --- 4. Initial Yield (Masked State) ---
|
388 |
+
initial_generated_tokens = x[0, prompt_length:].cpu()
|
389 |
+
vis_data_initial = []
|
390 |
+
for tok_id in initial_generated_tokens.tolist():
|
391 |
+
display_token = MASK_TOKEN
|
392 |
+
color = "#444444" # Dark Gray for masks
|
393 |
+
vis_data_initial.append((display_token, color))
|
394 |
+
|
395 |
+
previous_tokens_vis = initial_generated_tokens
|
396 |
+
yield history_copy, vis_data_initial, "" # Yield initial state
|
397 |
+
time.sleep(visualization_delay)
|
398 |
+
|
399 |
+
# --- 5. Step-by-Step Diffusion Loop ---
|
400 |
+
try:
|
401 |
+
start_time = time.time()
|
402 |
+
for i in range(steps):
|
403 |
+
# --- Model Forward Pass ---
|
404 |
+
mask_index = (x == MASK_ID) # Find masks in the *current* state x
|
405 |
+
if not mask_index.any(): # Stop if no masks left
|
406 |
+
print(f"No mask tokens left at step {i}. Stopping early.")
|
407 |
+
break
|
408 |
+
|
409 |
+
# print(f"Step {i}: Input shape {x.shape}, Mask sum {mask_index.sum()}") # Debug
|
410 |
+
# print(f"Step {i}: Input tokens (first/last 10): {x[0, :10].tolist()} ... {x[0, -10:].tolist()}") # Debug
|
411 |
+
|
412 |
+
# Call the model - ensure attention mask format is correct
|
413 |
+
# The model forward expects `attention_mask` usually of shape [B, N] or broadcastable.
|
414 |
+
# If we use `attention_mask_for_model = None`, it implies full attention.
|
415 |
+
# If we computed `attention_mask_for_model` as [B, 1, N, N], pass that.
|
416 |
+
# Let's try passing the [B, N] mask and let the model handle broadcasting/causality internally.
|
417 |
+
outputs = model(
|
418 |
+
input_ids=x,
|
419 |
+
attention_mask=full_attention_mask, # Pass the [B, N] mask
|
420 |
+
position_ids=None, # Let model compute default positions
|
421 |
+
use_cache=False, # No cache needed for diffusion steps
|
422 |
+
return_dict=True
|
423 |
+
)
|
424 |
+
logits = outputs.logits
|
425 |
+
|
426 |
+
# Shift logits like in original code? Check `generation_utils.py`.
|
427 |
+
# Yes, `logits = torch.cat([logits[:,:1], logits[:, :-1]], dim=1)`
|
428 |
+
# This seems to align logits with the *previous* token's prediction. Is this correct for diffusion?
|
429 |
+
# Let's assume the original code did this for a reason, perhaps related to how the model was trained or expects inputs.
|
430 |
+
# Update: Looking at standard LM forward pass, logits[t] predicts token[t+1].
|
431 |
+
# The shift aligns logits[t] with token[t]. Let's keep it.
|
432 |
+
logits = torch.cat([logits[:,:1], logits[:, :-1]], dim=1)
|
433 |
+
|
434 |
+
|
435 |
+
# Select logits for masked positions
|
436 |
+
# Ensure mask_index has the same batch dimension size as logits
|
437 |
+
# mask_index shape is [B, N], logits shape is [B, N, V]
|
438 |
+
# We need to select elements from the last dim of logits where mask is True
|
439 |
+
mask_logits = logits[mask_index] # This correctly selects [num_masked_tokens, V]
|
440 |
+
|
441 |
+
if mask_logits.numel() == 0: # If no masks, logits selection is empty
|
442 |
+
print(f"No masked tokens found for logit selection at step {i}. Stopping.")
|
443 |
+
break
|
444 |
+
|
445 |
+
# print(f"Step {i}: mask_logits shape: {mask_logits.shape}") # Debug
|
446 |
+
|
447 |
+
# --- Sampling / Remasking Logic ---
|
448 |
+
t = timesteps[i]
|
449 |
+
s = timesteps[i + 1]
|
450 |
+
|
451 |
+
x_new_masked_part = torch.full_like(x[mask_index], MASK_ID, device=device, dtype=torch.long)
|
452 |
+
|
453 |
+
if alg == 'origin':
|
454 |
+
# Original diffusion logic
|
455 |
+
p_transfer = (1.0 - s / t) if i < steps - 1 else 1.0 # Ensure float division
|
456 |
+
# Sample only for the tokens to be revealed in this step
|
457 |
+
num_masked = mask_logits.shape[0]
|
458 |
+
transfer_indices_relative = torch.rand(num_masked, device=device) < p_transfer
|
459 |
+
logits_to_sample = mask_logits[transfer_indices_relative]
|
460 |
+
|
461 |
+
if logits_to_sample.numel() > 0:
|
462 |
+
# print(f"Step {i} (origin): Sampling {logits_to_sample.shape[0]} tokens.") # Debug
|
463 |
+
_, sampled_tokens = sample_tokens(logits_to_sample, temperature=temperature, top_p=top_p_val, top_k=top_k_val)
|
464 |
+
# Place sampled tokens into the correct positions within the masked part
|
465 |
+
x_new_masked_part[transfer_indices_relative] = sampled_tokens
|
466 |
+
# else:
|
467 |
+
# print(f"Step {i} (origin): No tokens to sample (p_transfer={p_transfer}).") # Debug
|
468 |
+
|
469 |
+
else:
|
470 |
+
# Confidence-based algorithms (maskgit_plus, topk_margin, entropy)
|
471 |
+
use_margin = (alg == 'topk_margin')
|
472 |
+
use_entropy = (alg == 'entropy')
|
473 |
+
# print(f"Step {i} ({alg}): Sampling all {mask_logits.shape[0]} masked tokens for confidence.") # Debug
|
474 |
+
confidence, x0_candidates = sample_tokens(
|
475 |
+
mask_logits,
|
476 |
+
temperature=temperature,
|
477 |
+
top_p=top_p_val,
|
478 |
+
top_k=top_k_val,
|
479 |
+
margin_confidence=use_margin,
|
480 |
+
neg_entropy=use_entropy
|
481 |
+
)
|
482 |
+
# print(f"Step {i} ({alg}): Confidence range: [{confidence.min():.2f}, {confidence.max():.2f}]") # Debug
|
483 |
+
|
484 |
+
|
485 |
+
num_mask_token = mask_logits.shape[0]
|
486 |
+
# Calculate number to reveal based on time steps, ensure it's an int
|
487 |
+
target_num_revealed_float = num_mask_token * (1.0 - s / t)
|
488 |
+
number_transfer_tokens = int(target_num_revealed_float) if i < steps - 1 else num_mask_token
|
489 |
+
|
490 |
+
|
491 |
+
if number_transfer_tokens > 0:
|
492 |
+
# print(f"Step {i} ({alg}): Need to reveal {number_transfer_tokens} tokens.") # Debug
|
493 |
+
if alg_temp_val is None or alg_temp_val <= 0: # Use top-k confidence
|
494 |
+
# Sort by confidence (use negative entropy directly if alg='entropy')
|
495 |
+
# For entropy, lower (more negative) is higher confidence (less uncertainty)
|
496 |
+
sort_metric = confidence if alg != 'entropy' else -confidence
|
497 |
+
_, transfer_indices_relative = torch.topk(sort_metric, k=min(number_transfer_tokens, num_mask_token)) # Ensure k is not > num_mask_token
|
498 |
+
else: # Use sampling based on confidence temperature
|
499 |
+
conf_probs = confidence / alg_temp_val
|
500 |
+
# Check for inf/-inf before softmax
|
501 |
+
conf_probs = torch.nan_to_num(conf_probs, nan=0.0, posinf=1e9, neginf=-1e9)
|
502 |
+
conf_probs = F.softmax(conf_probs, dim=-1)
|
503 |
+
# Check probs sum to 1
|
504 |
+
if not torch.allclose(conf_probs.sum(), torch.tensor(1.0, device=device), atol=1e-4):
|
505 |
+
print(f"Warning step {i}: Confidence probabilities do not sum to 1 after softmax ({conf_probs.sum()}). Re-normalizing.")
|
506 |
+
conf_probs = conf_probs / conf_probs.sum(dim=-1, keepdim=True) # Normalize
|
507 |
+
|
508 |
+
# Ensure num_samples is valid
|
509 |
+
num_samples = min(number_transfer_tokens, num_mask_token)
|
510 |
+
if conf_probs.numel() > 0 and num_samples > 0:
|
511 |
+
try:
|
512 |
+
transfer_indices_relative = torch.multinomial(conf_probs, num_samples=num_samples, replacement=False)
|
513 |
+
except RuntimeError as e:
|
514 |
+
print(f"Warning step {i}: Multinomial sampling failed ('{e}'). Falling back to top-k.")
|
515 |
+
# Fallback to top-k if multinomial fails (e.g., due to prob issues)
|
516 |
+
sort_metric = confidence if alg != 'entropy' else -confidence
|
517 |
+
_, transfer_indices_relative = torch.topk(sort_metric, k=num_samples)
|
518 |
+
else:
|
519 |
+
transfer_indices_relative = torch.tensor([], dtype=torch.long, device=device) # No indices if no probs or num_samples=0
|
520 |
+
|
521 |
+
# Place the selected candidate tokens into the masked part update
|
522 |
+
if transfer_indices_relative.numel() > 0:
|
523 |
+
x_new_masked_part[transfer_indices_relative] = x0_candidates[transfer_indices_relative].clone()
|
524 |
+
# else:
|
525 |
+
# print(f"Step {i} ({alg}): No tokens revealed via confidence ({number_transfer_tokens} target).") # Debug
|
526 |
+
|
527 |
+
# Update the global state `x` only at the masked positions
|
528 |
+
x[mask_index] = x_new_masked_part
|
529 |
+
|
530 |
+
# --- Apply Constraints ---
|
531 |
+
# Constraints should be applied *after* sampling/revealing tokens for the step
|
532 |
+
x = apply_constraints_to_state(x, prompt_length, total_length, parsed_constraints, current_step=i)
|
533 |
+
|
534 |
+
# --- Yield Visualization ---
|
535 |
+
current_generated_tokens = x[0, prompt_length:].cpu() # Get generated part, move to CPU
|
536 |
+
vis_data = []
|
537 |
+
for j in range(gen_length):
|
538 |
+
current_tok_id = current_generated_tokens[j].item()
|
539 |
+
previous_tok_id = previous_tokens_vis[j].item() if previous_tokens_vis is not None else MASK_ID
|
540 |
|
541 |
+
try:
|
542 |
+
decoded_token = tokenizer.decode([current_tok_id], skip_special_tokens=False)
|
543 |
+
display_token = MASK_TOKEN if current_tok_id == MASK_ID else decoded_token
|
544 |
+
except Exception:
|
545 |
+
display_token = f"[ID:{current_tok_id}]" # Fallback
|
|
|
546 |
|
547 |
+
color = None
|
548 |
+
token_to_display = display_token
|
|
|
549 |
|
550 |
+
if current_tok_id == MASK_ID:
|
551 |
+
color = "#444444" # Dark Gray for masks
|
552 |
+
elif previous_tok_id == MASK_ID: # Token was just revealed
|
553 |
+
color = "#66CC66" # Light Green
|
554 |
+
else: # Token was already revealed
|
555 |
+
color = "#6699CC" # Light Blue
|
556 |
+
|
557 |
+
# Hide special tokens (PAD/EOS) if they were already revealed (LLaDA effect)
|
558 |
+
# Ensure PAD_ID and EOS_ID are not None before checking
|
559 |
+
should_hide = (PAD_ID is not None and current_tok_id == PAD_ID) or \
|
560 |
+
(EOS_ID is not None and current_tok_id == EOS_ID)
|
561 |
+
if should_hide and previous_tok_id == current_tok_id:
|
562 |
+
token_to_display = "" # Hide by making empty
|
563 |
+
color = None # No color for hidden
|
564 |
+
|
565 |
+
|
566 |
+
if token_to_display:
|
567 |
+
vis_data.append((token_to_display, color))
|
568 |
+
elif len(vis_data) > 0 and isinstance(vis_data[-1], tuple) and vis_data[-1][0] == " ":
|
569 |
+
# Avoid adding multiple spaces if tokens are hidden consecutively
|
570 |
+
pass
|
571 |
+
elif len(vis_data) > 0 and not isinstance(vis_data[-1], tuple) and vis_data[-1] == " ":
|
572 |
+
pass # Already added a space
|
573 |
+
elif len(vis_data) > 0 :
|
574 |
+
# Add a single space if hiding follows a visible token, improves readability slightly
|
575 |
+
# Let's simplify: just omit hidden tokens. Adding spaces might be complex.
|
576 |
+
pass
|
577 |
+
|
578 |
+
# Update previous state for the next iteration
|
579 |
+
previous_tokens_vis = current_generated_tokens
|
580 |
+
|
581 |
+
# Decode intermediate response (might be partial) - skip specials for readability
|
582 |
+
intermediate_response_tokens = x[0, prompt_length:]
|
583 |
+
intermediate_response_text = tokenizer.decode(
|
584 |
+
intermediate_response_tokens,
|
585 |
+
skip_special_tokens=True,
|
586 |
+
clean_up_tokenization_spaces=True
|
587 |
+
).strip()
|
588 |
+
|
589 |
+
# Yield current state
|
590 |
+
# We yield the *current* history, the vis data for this step, and intermediate text
|
591 |
+
# The final text will overwrite the intermediate text in the UI eventually
|
592 |
+
yield history_copy, vis_data, intermediate_response_text
|
593 |
+
time.sleep(visualization_delay)
|
594 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
595 |
end_time = time.time()
|
596 |
print(f"Dream generation finished in {end_time - start_time:.2f} seconds.")
|
597 |
|
598 |
+
# --- 6. Final Processing & Yield ---
|
599 |
+
final_sequence = x[0]
|
600 |
response_tokens = final_sequence[prompt_length:]
|
601 |
|
602 |
# Decode the final response text
|
|
|
606 |
clean_up_tokenization_spaces=True
|
607 |
).strip()
|
608 |
|
609 |
+
# Update history with the final response *before* the last yield
|
610 |
+
history_copy[-1][1] = final_response_text
|
611 |
|
612 |
+
# Yield the final state (which might be the same as the last yielded state if loop finished)
|
613 |
+
# Need to format vis_data one last time based on the final `x`
|
614 |
+
final_generated_tokens = x[0, prompt_length:].cpu()
|
615 |
+
vis_data_final = []
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
616 |
for j in range(gen_length):
|
617 |
+
current_tok_id = final_generated_tokens[j].item()
|
618 |
previous_tok_id = previous_tokens_vis[j].item() if previous_tokens_vis is not None else MASK_ID
|
619 |
|
|
|
620 |
try:
|
|
|
621 |
decoded_token = tokenizer.decode([current_tok_id], skip_special_tokens=False)
|
622 |
+
display_token = MASK_TOKEN if current_tok_id == MASK_ID else decoded_token
|
|
|
|
|
|
|
|
|
|
|
623 |
except Exception:
|
624 |
display_token = f"[ID:{current_tok_id}]" # Fallback
|
625 |
|
|
|
626 |
color = None
|
627 |
token_to_display = display_token
|
628 |
|
629 |
if current_tok_id == MASK_ID:
|
630 |
+
color = "#444444"
|
631 |
+
elif previous_tok_id == MASK_ID:
|
632 |
+
color = "#66CC66"
|
633 |
+
else:
|
634 |
+
color = "#6699CC"
|
635 |
+
|
636 |
+
should_hide = (PAD_ID is not None and current_tok_id == PAD_ID) or \
|
637 |
+
(EOS_ID is not None and current_tok_id == EOS_ID)
|
638 |
+
if should_hide and previous_tok_id == current_tok_id:
|
|
|
639 |
token_to_display = ""
|
640 |
+
color = None
|
641 |
|
642 |
+
if token_to_display:
|
643 |
+
vis_data_final.append((token_to_display, color))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
644 |
|
645 |
+
# Yield the final history, final visualization, and final text
|
646 |
+
yield history_copy, vis_data_final, final_response_text
|
647 |
+
print("Visualization streaming complete.")
|
648 |
|
|
|
|
|
649 |
|
650 |
+
except Exception as e:
|
651 |
+
print(f"Error during generation or processing: {e}")
|
652 |
+
import traceback
|
653 |
+
traceback.print_exc()
|
654 |
+
# Update history with error message? Or leave as None? Let's leave as None.
|
655 |
+
yield history_copy, [("Error during generation.", "red")], ""
|
656 |
+
return
|
657 |
|
658 |
|
659 |
+
# --- Gradio UI (Remains largely the same, ensures outputs match yield structure) ---
|
660 |
css = '''
|
661 |
.category-legend{display:none}
|
662 |
button{min-height: 60px}
|
|
|
666 |
gr.Markdown("# Dream 7B - Diffusion Language Model Demo")
|
667 |
gr.Markdown(
|
668 |
"[[Model Card](https://huggingface.co/Dream-org/Dream-v0-Instruct-7B)] "
|
669 |
+
"[[Blog](https://hkunlp.github.io/blog/2025/dream/)]" # Note: Link might be hypothetical
|
670 |
)
|
671 |
|
672 |
# STATE MANAGEMENT
|
673 |
+
# chat_history = gr.State([]) # Use gr.Chatbot's internal state implicitly if possible, or manage manually
|
674 |
+
# Let's manage manually with a list for clarity with yielding updates
|
675 |
+
_chat_history_store = gr.State([]) # Hidden state to store actual history list
|
676 |
|
677 |
# UI COMPONENTS
|
678 |
with gr.Row():
|
|
|
681 |
label="Conversation",
|
682 |
height=500,
|
683 |
show_copy_button=True,
|
684 |
+
bubble_full_width=False,
|
685 |
+
# value=[] # Initialize chatbot UI empty
|
686 |
)
|
687 |
|
688 |
# Message input
|
|
|
709 |
output_vis = gr.HighlightedText(
|
710 |
label="Denoising Process Visualization",
|
711 |
combine_adjacent=False,
|
712 |
+
show_legend=True, # Legend isn't very informative here
|
713 |
+
interactive=False # Not interactive
|
714 |
+
)
|
715 |
+
# Add a text box to display the final/intermediate response clearly
|
716 |
+
response_text_display = gr.Textbox(
|
717 |
+
label="Generated Response",
|
718 |
+
interactive=False,
|
719 |
+
lines=5 # Show a few lines
|
720 |
)
|
721 |
|
722 |
+
|
723 |
# Advanced generation settings
|
724 |
with gr.Accordion("Generation Settings", open=False):
|
725 |
with gr.Row():
|
|
|
734 |
with gr.Row():
|
735 |
temperature = gr.Slider(
|
736 |
minimum=0.0, maximum=1.0, value=0.4, step=0.05,
|
737 |
+
label="Temperature (0 = greedy)"
|
738 |
)
|
739 |
alg_temp = gr.Slider(
|
740 |
minimum=0.0, maximum=1.0, value=0.1, step=0.05,
|
741 |
+
label="Remasking Temp (Confidence Algs)"
|
742 |
)
|
743 |
|
744 |
with gr.Row():
|
745 |
top_p = gr.Slider(
|
746 |
minimum=0.0, maximum=1.0, value=0.95, step=0.05,
|
747 |
+
label="Top-P (<=0 or >=1 disables)" # Clarify disabling condition
|
748 |
)
|
749 |
top_k = gr.Slider(
|
750 |
minimum=0, maximum=200, value=0, step=5,
|
751 |
+
label="Top-K (0 disables)"
|
752 |
)
|
753 |
|
754 |
with gr.Row():
|
|
|
760 |
|
761 |
with gr.Row():
|
762 |
visualization_delay = gr.Slider(
|
763 |
+
minimum=0.0, maximum=0.5, value=0.03, step=0.01, # Slightly faster default
|
764 |
label="Visualization Delay (seconds)"
|
765 |
)
|
766 |
|
767 |
# Clear button
|
768 |
clear_btn = gr.Button("Clear Conversation")
|
769 |
|
|
|
|
|
|
|
770 |
# --- Event Handlers ---
|
771 |
|
772 |
+
def add_user_message_to_history(message: str, history_store: List[List[Optional[str]]]):
|
773 |
"""Adds user message, clears input, prepares for bot response."""
|
774 |
if not message.strip():
|
775 |
gr.Warning("Please enter a message.")
|
776 |
+
# Return unchanged history, empty vis, empty response text
|
777 |
+
return history_store, history_store, "", [], ""
|
778 |
|
779 |
# Add user message with placeholder for bot response
|
780 |
+
history_store.append([message, None])
|
781 |
+
# Return updated history store, history for chatbot UI, empty input, empty vis, empty response
|
782 |
+
return history_store, history_store, "", [], ""
|
|
|
783 |
|
784 |
def clear_conversation():
|
785 |
+
"""Clears the chat history, visualization, and response text."""
|
786 |
+
return [], [], "", [], "" # History store, chatbot UI, input, vis, response text
|
787 |
|
788 |
# --- Connect UI elements ---
|
789 |
|
790 |
# Define the inputs for the generation function once
|
791 |
generation_inputs = [
|
792 |
+
_chat_history_store, gen_length, steps, constraints_input,
|
793 |
temperature, top_p, top_k, remasking_strategy, alg_temp,
|
794 |
visualization_delay
|
795 |
]
|
796 |
# Define the outputs for the generation function
|
797 |
+
# Now yields: history_copy, vis_data, intermediate_response_text
|
798 |
+
# Map these to: chatbot_ui, output_vis, response_text_display
|
799 |
+
generation_outputs = [chatbot_ui, output_vis, response_text_display]
|
800 |
|
801 |
# Handle Textbox Submission (Enter key)
|
802 |
submit_listener = user_input.submit(
|
803 |
fn=add_user_message_to_history,
|
804 |
+
inputs=[user_input, _chat_history_store],
|
805 |
+
outputs=[_chat_history_store, chatbot_ui, user_input, output_vis, response_text_display] # Step 1: Add user msg & clear outputs
|
806 |
)
|
807 |
# Chain the bot response generation after the user message is added
|
808 |
submit_listener.then(
|
809 |
fn=generate_dream_response,
|
810 |
inputs=generation_inputs,
|
811 |
+
outputs=generation_outputs, # Step 2: Generate response and stream vis/text
|
812 |
+
show_progress="hidden" # Hide default progress bar as we have live vis
|
813 |
)
|
814 |
|
815 |
# Handle Send Button Click
|
816 |
click_listener = send_btn.click(
|
817 |
fn=add_user_message_to_history,
|
818 |
+
inputs=[user_input, _chat_history_store],
|
819 |
+
outputs=[_chat_history_store, chatbot_ui, user_input, output_vis, response_text_display] # Step 1: Add user msg & clear outputs
|
820 |
)
|
821 |
# Chain the bot response generation after the user message is added
|
822 |
click_listener.then(
|
823 |
fn=generate_dream_response,
|
824 |
inputs=generation_inputs,
|
825 |
+
outputs=generation_outputs, # Step 2: Generate response and stream vis/text
|
826 |
+
show_progress="hidden"
|
827 |
)
|
828 |
|
829 |
+
# Clear Button Action
|
830 |
clear_btn.click(
|
831 |
clear_conversation,
|
832 |
inputs=[],
|
833 |
+
outputs=[_chat_history_store, chatbot_ui, user_input, output_vis, response_text_display]
|
834 |
)
|
835 |
|
836 |
return demo
|
|
|
839 |
if __name__ == "__main__":
|
840 |
demo = create_chatbot_demo()
|
841 |
# Use queue for handling multiple users and streaming
|
842 |
+
demo.queue().launch(debug=True, share=False) # Set share=True for public link
|