Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
|
@@ -68,8 +68,6 @@ def generate_response_with_visualization(model, tokenizer, device, messages, gen
|
|
| 68 |
Returns:
|
| 69 |
List of visualization states showing the progression and final text
|
| 70 |
"""
|
| 71 |
-
# Set random seed for reproducibility
|
| 72 |
-
torch.manual_seed(42)
|
| 73 |
|
| 74 |
# Process constraints
|
| 75 |
if constraints is None:
|
|
@@ -212,29 +210,12 @@ def generate_response_with_visualization(model, tokenizer, device, messages, gen
|
|
| 212 |
response_text = tokenizer.decode(response_tokens, skip_special_tokens=True)
|
| 213 |
|
| 214 |
# Clean the response text
|
| 215 |
-
final_text =
|
|
|
|
|
|
|
| 216 |
|
| 217 |
return visualization_states, final_text
|
| 218 |
|
| 219 |
-
def clean_output_text(text):
|
| 220 |
-
"""Clean the output text to remove special tokens and fix spacing"""
|
| 221 |
-
# Remove any remaining [MASK] tokens
|
| 222 |
-
text = text.replace(MASK_TOKEN, "")
|
| 223 |
-
|
| 224 |
-
# Fix common spacing issues with tokenization
|
| 225 |
-
text = re.sub(r'\s+', ' ', text) # Remove multiple spaces
|
| 226 |
-
text = re.sub(r' \.', '.', text) # Fix spacing before periods
|
| 227 |
-
text = re.sub(r' ,', ',', text) # Fix spacing before commas
|
| 228 |
-
text = re.sub(r' !', '!', text) # Fix spacing before exclamation marks
|
| 229 |
-
text = re.sub(r' \?', '?', text) # Fix spacing before question marks
|
| 230 |
-
text = re.sub(r' ;', ';', text) # Fix spacing before semicolons
|
| 231 |
-
text = re.sub(r' :', ':', text) # Fix spacing before colons
|
| 232 |
-
|
| 233 |
-
# Fix beginning and end spacing
|
| 234 |
-
text = text.strip()
|
| 235 |
-
|
| 236 |
-
return text
|
| 237 |
-
|
| 238 |
css = '''
|
| 239 |
.category-legend{display:none}
|
| 240 |
'''
|
|
|
|
| 68 |
Returns:
|
| 69 |
List of visualization states showing the progression and final text
|
| 70 |
"""
|
|
|
|
|
|
|
| 71 |
|
| 72 |
# Process constraints
|
| 73 |
if constraints is None:
|
|
|
|
| 210 |
response_text = tokenizer.decode(response_tokens, skip_special_tokens=True)
|
| 211 |
|
| 212 |
# Clean the response text
|
| 213 |
+
final_text = tokenizer.decode(response_tokens,
|
| 214 |
+
skip_special_tokens=True,
|
| 215 |
+
clean_up_tokenization_spaces=True)
|
| 216 |
|
| 217 |
return visualization_states, final_text
|
| 218 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 219 |
css = '''
|
| 220 |
.category-legend{display:none}
|
| 221 |
'''
|