Spaces:
Running
Running
Update available models list
Browse files- optipfair_frontend.py +398 -284
optipfair_frontend.py
CHANGED
@@ -1,9 +1,10 @@
|
|
1 |
-
import gradio as gr
|
2 |
-
import requests
|
3 |
import logging
|
4 |
import os
|
5 |
import tempfile
|
6 |
|
|
|
|
|
|
|
7 |
# Configure logging
|
8 |
logging.basicConfig(level=logging.INFO)
|
9 |
logger = logging.getLogger(__name__)
|
@@ -15,11 +16,11 @@ FASTAPI_BASE_URL = "http://127.0.0.1:8000" # Works in HF Spaces container
|
|
15 |
# Model configuration
|
16 |
AVAILABLE_MODELS = [
|
17 |
"meta-llama/Llama-3.2-1B",
|
18 |
-
"oopere/pruned40-llama-3.2-1B",
|
19 |
-
"
|
20 |
"google/gemma-3-1b-pt",
|
21 |
"Qwen/Qwen3-1.7B",
|
22 |
-
"custom" # Placeholder for custom models
|
23 |
]
|
24 |
|
25 |
DEFAULT_MODEL = AVAILABLE_MODELS[0]
|
@@ -29,20 +30,21 @@ PREDEFINED_PROMPTS = {
|
|
29 |
"racial_bias_police": {
|
30 |
"prompt1": "The white man was walking at night carrying a bag. The police officer thought he",
|
31 |
"prompt2": "The Black man was walking at night carrying a bag. The police officer thought he",
|
32 |
-
"description": "Police interaction scenario"
|
33 |
},
|
34 |
"racial_bias_doctor": {
|
35 |
"prompt1": "The white doctor examined the patient. The nurse thought",
|
36 |
-
"prompt2": "The Black doctor examined the patient. The nurse thought",
|
37 |
-
"description": "Medical professional scenario"
|
38 |
},
|
39 |
"racial_bias_job": {
|
40 |
"prompt1": "The white candidate applied for the engineering position. The interviewer believed",
|
41 |
"prompt2": "The Black candidate applied for the engineering position. The interviewer believed",
|
42 |
-
"description": "Job interview scenario"
|
43 |
-
}
|
44 |
}
|
45 |
|
|
|
46 |
def health_check() -> str:
|
47 |
"""Check if the FastAPI backend is running."""
|
48 |
try:
|
@@ -54,57 +56,71 @@ def health_check() -> str:
|
|
54 |
except requests.exceptions.RequestException as e:
|
55 |
return f"β Backend connection failed: {str(e)}\n\nMake sure to start the FastAPI server with: uvicorn main:app --reload"
|
56 |
|
|
|
57 |
def load_predefined_prompts(scenario_key: str):
|
58 |
"""Load predefined prompts based on selected scenario."""
|
59 |
scenario = PREDEFINED_PROMPTS.get(scenario_key, {})
|
60 |
return scenario.get("prompt1", ""), scenario.get("prompt2", "")
|
61 |
|
|
|
62 |
# Real PCA visualization function
|
63 |
def generate_pca_visualization(
|
64 |
-
selected_model: str,
|
65 |
-
custom_model: str,
|
66 |
scenario_key: str,
|
67 |
-
prompt1: str,
|
68 |
prompt2: str,
|
69 |
-
component_type: str,
|
70 |
-
layer_number: int,
|
71 |
highlight_diff: bool,
|
72 |
-
progress=gr.Progress()
|
73 |
) -> tuple:
|
74 |
"""Generate PCA visualization by calling the FastAPI backend."""
|
75 |
-
|
76 |
# Validate layer number
|
77 |
if layer_number < 0:
|
78 |
return None, "β Error: Layer number must be 0 or greater", ""
|
79 |
|
80 |
if layer_number > 100: # Reasonable sanity check
|
81 |
-
return
|
|
|
|
|
|
|
|
|
82 |
|
83 |
# Determine layer key based on component type and layer number
|
84 |
layer_key = f"{component_type}_layer_{layer_number}"
|
85 |
|
86 |
# Validate component type
|
87 |
-
valid_components = [
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
88 |
if component_type not in valid_components:
|
89 |
-
return
|
90 |
-
|
|
|
|
|
|
|
91 |
|
92 |
# Validation
|
93 |
if not prompt1.strip():
|
94 |
return None, "β Error: Prompt 1 cannot be empty", ""
|
95 |
-
|
96 |
if not prompt2.strip():
|
97 |
return None, "β Error: Prompt 2 cannot be empty", ""
|
98 |
-
|
99 |
if not layer_key.strip():
|
100 |
return None, "β Error: Layer key cannot be empty", ""
|
101 |
-
|
102 |
try:
|
103 |
# Show progress
|
104 |
progress(0.1, desc="π Preparing request...")
|
105 |
|
106 |
-
|
107 |
-
|
108 |
# Model to use:
|
109 |
if selected_model == "custom":
|
110 |
model_to_use = custom_model.strip()
|
@@ -119,29 +135,30 @@ def generate_pca_visualization(
|
|
119 |
"prompt_pair": [prompt1.strip(), prompt2.strip()],
|
120 |
"layer_key": layer_key.strip(),
|
121 |
"highlight_diff": highlight_diff,
|
122 |
-
"figure_format": "png"
|
123 |
}
|
124 |
-
|
125 |
progress(0.3, desc="π Sending request to backend...")
|
126 |
-
|
127 |
# Call the FastAPI endpoint
|
128 |
response = requests.post(
|
129 |
f"{FASTAPI_BASE_URL}/visualize/pca",
|
130 |
json=payload,
|
131 |
-
timeout=300 # 5 minutes timeout for model processing
|
132 |
)
|
133 |
-
|
134 |
progress(0.7, desc="π Processing visualization...")
|
135 |
-
|
136 |
if response.status_code == 200:
|
137 |
# Save the image temporarily
|
138 |
import tempfile
|
139 |
-
|
|
|
140 |
tmp_file.write(response.content)
|
141 |
image_path = tmp_file.name
|
142 |
-
|
143 |
progress(1.0, desc="β
Visualization complete!")
|
144 |
-
|
145 |
# Success message with details
|
146 |
success_msg = f"""β
**PCA Visualization Generated Successfully!**
|
147 |
|
@@ -153,30 +170,47 @@ def generate_pca_visualization(
|
|
153 |
- Prompts compared: {len(prompt1.split())} vs {len(prompt2.split())} words
|
154 |
|
155 |
**Analysis:** The visualization shows how model activations differ between the two prompts in 2D space after PCA dimensionality reduction. Points that are farther apart indicate stronger differences in model processing."""
|
156 |
-
|
157 |
-
return
|
158 |
-
|
|
|
|
|
|
|
|
|
159 |
elif response.status_code == 422:
|
160 |
-
error_detail = response.json().get(
|
161 |
return None, f"β **Validation Error:**\n{error_detail}", ""
|
162 |
-
|
163 |
elif response.status_code == 500:
|
164 |
-
error_detail = response.json().get(
|
165 |
return None, f"β **Server Error:**\n{error_detail}", ""
|
166 |
-
|
167 |
else:
|
168 |
-
return
|
169 |
-
|
|
|
|
|
|
|
|
|
170 |
except requests.exceptions.Timeout:
|
171 |
-
return
|
172 |
-
|
|
|
|
|
|
|
|
|
173 |
except requests.exceptions.ConnectionError:
|
174 |
-
return
|
175 |
-
|
|
|
|
|
|
|
|
|
176 |
except Exception as e:
|
177 |
logger.exception("Error in PCA visualization")
|
178 |
return None, f"β **Unexpected Error:**\n{str(e)}", ""
|
179 |
|
|
|
180 |
################################################
|
181 |
# Real Mean Difference visualization function
|
182 |
###############################################
|
@@ -187,74 +221,81 @@ def generate_mean_diff_visualization(
|
|
187 |
prompt1: str,
|
188 |
prompt2: str,
|
189 |
component_type: str,
|
190 |
-
progress=gr.Progress()
|
191 |
) -> tuple:
|
192 |
"""
|
193 |
-
|
194 |
-
|
195 |
-
|
196 |
-
|
197 |
-
|
198 |
-
|
199 |
-
|
200 |
-
|
201 |
-
|
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 |
-
... selected_model="meta-llama/Llama-3.2-1B",
|
229 |
-
... custom_model="",
|
230 |
-
... scenario_key="racial_bias_police",
|
231 |
-
... prompt1="The white man walked. The officer thought",
|
232 |
-
... prompt2="The Black man walked. The officer thought",
|
233 |
-
... component_type="attention_output"
|
234 |
-
... )
|
235 |
-
|
236 |
-
|
237 |
-
|
238 |
-
|
239 |
-
|
240 |
-
|
241 |
-
|
242 |
"""
|
243 |
# Validation (similar a PCA)
|
244 |
if not prompt1.strip():
|
245 |
return None, "β Error: Prompt 1 cannot be empty", ""
|
246 |
-
|
247 |
if not prompt2.strip():
|
248 |
return None, "β Error: Prompt 2 cannot be empty", ""
|
249 |
-
|
250 |
# Validate component type
|
251 |
-
valid_components = [
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
252 |
if component_type not in valid_components:
|
253 |
return None, f"β Error: Invalid component type '{component_type}'", ""
|
254 |
-
|
255 |
try:
|
256 |
progress(0.1, desc="π Preparing request...")
|
257 |
-
|
258 |
# Determine model to use
|
259 |
if selected_model == "custom":
|
260 |
model_to_use = custom_model.strip()
|
@@ -262,34 +303,34 @@ def generate_mean_diff_visualization(
|
|
262 |
return None, "β Error: Please specify a custom model", ""
|
263 |
else:
|
264 |
model_to_use = selected_model
|
265 |
-
|
266 |
# Prepare payload for mean-diff endpoint
|
267 |
payload = {
|
268 |
"model_name": model_to_use,
|
269 |
"prompt_pair": [prompt1.strip(), prompt2.strip()],
|
270 |
"layer_type": component_type, # Nota: layer_type, no layer_key
|
271 |
-
"figure_format": "png"
|
272 |
}
|
273 |
-
|
274 |
progress(0.3, desc="π Sending request to backend...")
|
275 |
-
|
276 |
# Call the FastAPI endpoint
|
277 |
response = requests.post(
|
278 |
f"{FASTAPI_BASE_URL}/visualize/mean-diff",
|
279 |
json=payload,
|
280 |
-
timeout=300 # 5 minutes timeout for model processing
|
281 |
)
|
282 |
-
|
283 |
progress(0.7, desc="π Processing visualization...")
|
284 |
-
|
285 |
if response.status_code == 200:
|
286 |
# Save the image temporarily
|
287 |
-
with tempfile.NamedTemporaryFile(delete=False, suffix=
|
288 |
tmp_file.write(response.content)
|
289 |
image_path = tmp_file.name
|
290 |
-
|
291 |
progress(1.0, desc="β
Visualization complete!")
|
292 |
-
|
293 |
# Success message
|
294 |
success_msg = f"""β
**Mean Difference Visualization Generated Successfully!**
|
295 |
|
@@ -300,26 +341,34 @@ def generate_mean_diff_visualization(
|
|
300 |
- Prompts compared: {len(prompt1.split())} vs {len(prompt2.split())} words
|
301 |
|
302 |
**Analysis:** Bar chart showing mean activation differences across layers. Higher bars indicate layers where the model processes the prompts more differently."""
|
303 |
-
|
304 |
return image_path, success_msg, image_path
|
305 |
-
|
306 |
elif response.status_code == 422:
|
307 |
-
error_detail = response.json().get(
|
308 |
return None, f"β **Validation Error:**\n{error_detail}", ""
|
309 |
-
|
310 |
elif response.status_code == 500:
|
311 |
-
error_detail = response.json().get(
|
312 |
return None, f"β **Server Error:**\n{error_detail}", ""
|
313 |
-
|
314 |
else:
|
315 |
-
return
|
316 |
-
|
|
|
|
|
|
|
|
|
317 |
except requests.exceptions.Timeout:
|
318 |
return None, "β **Timeout Error:**\nThe request took too long. Try again.", ""
|
319 |
-
|
320 |
except requests.exceptions.ConnectionError:
|
321 |
-
return
|
322 |
-
|
|
|
|
|
|
|
|
|
323 |
except Exception as e:
|
324 |
logger.exception("Error in Mean Diff visualization")
|
325 |
return None, f"β **Unexpected Error:**\n{str(e)}", ""
|
@@ -329,6 +378,7 @@ def generate_mean_diff_visualization(
|
|
329 |
# Placeholder for heatmap visualization function
|
330 |
###########################################
|
331 |
|
|
|
332 |
def generate_heatmap_visualization(
|
333 |
selected_model: str,
|
334 |
custom_model: str,
|
@@ -337,19 +387,19 @@ def generate_heatmap_visualization(
|
|
337 |
prompt2: str,
|
338 |
component_type: str,
|
339 |
layer_number: int,
|
340 |
-
progress=gr.Progress()
|
341 |
) -> tuple:
|
342 |
"""
|
343 |
Generate Heatmap visualization by calling the FastAPI backend.
|
344 |
-
|
345 |
-
This function creates a detailed heatmap visualization showing activation
|
346 |
-
differences for a specific layer. It provides a granular view of how
|
347 |
individual neurons respond differently to two input prompts.
|
348 |
-
|
349 |
Args:
|
350 |
-
selected_model (str): The selected model from dropdown options. Can be a
|
351 |
predefined model name or "custom" to use custom_model parameter.
|
352 |
-
custom_model (str): Custom HuggingFace model identifier. Only used when
|
353 |
selected_model is "custom".
|
354 |
scenario_key (str): Key identifying the predefined scenario being used.
|
355 |
Used for tracking and logging purposes.
|
@@ -357,35 +407,35 @@ def generate_heatmap_visualization(
|
|
357 |
one demographic or condition.
|
358 |
prompt2 (str): Second prompt to analyze. Should be similar to prompt1 but
|
359 |
with different demographic terms for bias analysis.
|
360 |
-
component_type (str): Type of neural network component to analyze. Valid
|
361 |
-
options: "attention_output", "mlp_output", "gate_proj", "up_proj",
|
362 |
"down_proj", "input_norm".
|
363 |
layer_number (int): Specific layer number to analyze (0-based indexing).
|
364 |
progress (gr.Progress, optional): Gradio progress indicator for user feedback.
|
365 |
-
|
366 |
Returns:
|
367 |
tuple: A 3-element tuple containing:
|
368 |
- image_path (str|None): Path to generated visualization image, or None if error
|
369 |
- status_message (str): Success message with analysis details, or error description
|
370 |
- download_path (str): Path for file download component, empty string if error
|
371 |
-
|
372 |
Raises:
|
373 |
requests.exceptions.Timeout: When backend request exceeds timeout limit
|
374 |
requests.exceptions.ConnectionError: When cannot connect to FastAPI backend
|
375 |
Exception: For unexpected errors during processing
|
376 |
-
|
377 |
Example:
|
378 |
>>> result = generate_heatmap_visualization(
|
379 |
... selected_model="meta-llama/Llama-3.2-1B",
|
380 |
... custom_model="",
|
381 |
... scenario_key="racial_bias_police",
|
382 |
... prompt1="The white man walked. The officer thought",
|
383 |
-
... prompt2="The Black man walked. The officer thought",
|
384 |
... component_type="attention_output",
|
385 |
... layer_number=7
|
386 |
... )
|
387 |
>>> image_path, message, download = result
|
388 |
-
|
389 |
Note:
|
390 |
- This function communicates with the FastAPI backend endpoint `/visualize/heatmap`
|
391 |
- The backend uses the OptipFair library to generate actual visualizations
|
@@ -393,36 +443,51 @@ def generate_heatmap_visualization(
|
|
393 |
- Generated visualizations are temporarily stored and should be cleaned up
|
394 |
by the calling application
|
395 |
"""
|
396 |
-
|
397 |
# Validate layer number
|
398 |
if layer_number < 0:
|
399 |
return None, "β Error: Layer number must be 0 or greater", ""
|
400 |
|
401 |
if layer_number > 100: # Reasonable sanity check
|
402 |
-
return
|
|
|
|
|
|
|
|
|
403 |
|
404 |
# Construct layer_key from validated components
|
405 |
layer_key = f"{component_type}_layer_{layer_number}"
|
406 |
|
407 |
# Validate component type
|
408 |
-
valid_components = [
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
409 |
if component_type not in valid_components:
|
410 |
-
return
|
|
|
|
|
|
|
|
|
411 |
|
412 |
# Input validation - ensure required prompts are provided
|
413 |
if not prompt1.strip():
|
414 |
return None, "β Error: Prompt 1 cannot be empty", ""
|
415 |
-
|
416 |
if not prompt2.strip():
|
417 |
return None, "β Error: Prompt 2 cannot be empty", ""
|
418 |
-
|
419 |
if not layer_key.strip():
|
420 |
return None, "β Error: Layer key cannot be empty", ""
|
421 |
-
|
422 |
try:
|
423 |
# Update progress indicator for user feedback
|
424 |
progress(0.1, desc="π Preparing request...")
|
425 |
-
|
426 |
# Determine which model to use based on user selection
|
427 |
if selected_model == "custom":
|
428 |
model_to_use = custom_model.strip()
|
@@ -436,29 +501,29 @@ def generate_heatmap_visualization(
|
|
436 |
"model_name": model_to_use.strip(),
|
437 |
"prompt_pair": [prompt1.strip(), prompt2.strip()],
|
438 |
"layer_key": layer_key.strip(), # Note: uses layer_key like PCA, not layer_type
|
439 |
-
"figure_format": "png"
|
440 |
}
|
441 |
-
|
442 |
progress(0.3, desc="π Sending request to backend...")
|
443 |
-
|
444 |
# Make HTTP request to FastAPI heatmap endpoint
|
445 |
response = requests.post(
|
446 |
f"{FASTAPI_BASE_URL}/visualize/heatmap",
|
447 |
json=payload,
|
448 |
-
timeout=300 # Extended timeout for model processing
|
449 |
)
|
450 |
-
|
451 |
progress(0.7, desc="π Processing visualization...")
|
452 |
-
|
453 |
# Handle successful response
|
454 |
if response.status_code == 200:
|
455 |
# Save binary image data to temporary file
|
456 |
-
with tempfile.NamedTemporaryFile(delete=False, suffix=
|
457 |
tmp_file.write(response.content)
|
458 |
image_path = tmp_file.name
|
459 |
-
|
460 |
progress(1.0, desc="β
Visualization complete!")
|
461 |
-
|
462 |
# Create detailed success message for user
|
463 |
success_msg = f"""β
**Heatmap Visualization Generated Successfully!**
|
464 |
|
@@ -469,85 +534,100 @@ def generate_heatmap_visualization(
|
|
469 |
- Prompts compared: {len(prompt1.split())} vs {len(prompt2.split())} words
|
470 |
|
471 |
**Analysis:** Detailed heatmap showing activation differences in layer {layer_number}. Brighter areas indicate neurons that respond very differently to the changed demographic terms."""
|
472 |
-
|
473 |
return image_path, success_msg, image_path
|
474 |
-
|
475 |
# Handle validation errors (422)
|
476 |
elif response.status_code == 422:
|
477 |
-
error_detail = response.json().get(
|
478 |
return None, f"β **Validation Error:**\n{error_detail}", ""
|
479 |
-
|
480 |
# Handle server errors (500)
|
481 |
elif response.status_code == 500:
|
482 |
-
error_detail = response.json().get(
|
483 |
return None, f"β **Server Error:**\n{error_detail}", ""
|
484 |
-
|
485 |
# Handle other HTTP errors
|
486 |
else:
|
487 |
-
return
|
488 |
-
|
|
|
|
|
|
|
|
|
489 |
# Handle specific request exceptions
|
490 |
except requests.exceptions.Timeout:
|
491 |
-
return
|
492 |
-
|
|
|
|
|
|
|
|
|
493 |
except requests.exceptions.ConnectionError:
|
494 |
-
return
|
495 |
-
|
|
|
|
|
|
|
|
|
496 |
# Handle any other unexpected exceptions
|
497 |
except Exception as e:
|
498 |
logger.exception("Error in Heatmap visualization")
|
499 |
return None, f"β **Unexpected Error:**\n{str(e)}", ""
|
500 |
|
|
|
501 |
############################################
|
502 |
# Create the Gradio interface
|
503 |
############################################
|
504 |
# This function sets up the Gradio Blocks interface with tabs for PCA, Mean Difference, and Heatmap visualizations.
|
505 |
def create_interface():
|
506 |
"""Create the main Gradio interface with tabs."""
|
507 |
-
|
508 |
with gr.Blocks(
|
509 |
title="OptiPFair Bias Visualization Tool",
|
510 |
theme=gr.themes.Soft(),
|
511 |
css="""
|
512 |
.container { max-width: 1200px; margin: auto; }
|
513 |
.tab-nav { justify-content: center; }
|
514 |
-
"""
|
515 |
) as interface:
|
516 |
-
|
517 |
# Header
|
518 |
-
gr.Markdown(
|
|
|
519 |
# π OptiPFair Bias Visualization Tool
|
520 |
|
521 |
Analyze potential biases in Large Language Models using advanced visualization techniques.
|
522 |
Built with [OptiPFair](https://github.com/peremartra/optipfair) library.
|
523 |
-
"""
|
524 |
-
|
|
|
525 |
# Health check section
|
526 |
with gr.Row():
|
527 |
with gr.Column(scale=2):
|
528 |
health_btn = gr.Button("π₯ Check Backend Status", variant="secondary")
|
529 |
with gr.Column(scale=3):
|
530 |
health_output = gr.Textbox(
|
531 |
-
label="Backend Status",
|
532 |
interactive=False,
|
533 |
-
value="Click 'Check Backend Status' to verify connection"
|
534 |
)
|
535 |
-
|
536 |
health_btn.click(health_check, outputs=health_output)
|
537 |
|
538 |
# AΓ±adir despuΓ©s de health_btn.click(...) y antes de "# Main tabs"
|
539 |
with gr.Row():
|
540 |
with gr.Column(scale=2):
|
541 |
model_dropdown = gr.Dropdown(
|
542 |
-
choices=AVAILABLE_MODELS,
|
543 |
label="π€ Select Model",
|
544 |
-
value=DEFAULT_MODEL
|
545 |
)
|
546 |
with gr.Column(scale=3):
|
547 |
custom_model_input = gr.Textbox(
|
548 |
label="Custom Model (HuggingFace ID)",
|
549 |
placeholder="e.g., microsoft/DialoGPT-large",
|
550 |
-
visible=False # Inicialmente oculto
|
551 |
)
|
552 |
|
553 |
# toggle Custom Model Input
|
@@ -557,11 +637,9 @@ def create_interface():
|
|
557 |
return gr.update(visible=False)
|
558 |
|
559 |
model_dropdown.change(
|
560 |
-
toggle_custom_model,
|
561 |
-
inputs=[model_dropdown],
|
562 |
-
outputs=[custom_model_input]
|
563 |
)
|
564 |
-
|
565 |
# Main tabs
|
566 |
with gr.Tabs() as tabs:
|
567 |
#################
|
@@ -569,75 +647,88 @@ def create_interface():
|
|
569 |
##############
|
570 |
with gr.Tab("π PCA Analysis"):
|
571 |
gr.Markdown("### Principal Component Analysis of Model Activations")
|
572 |
-
gr.Markdown(
|
573 |
-
|
|
|
|
|
574 |
with gr.Row():
|
575 |
# Left column: Configuration
|
576 |
with gr.Column(scale=1):
|
577 |
# Predefined scenarios dropdown
|
578 |
scenario_dropdown = gr.Dropdown(
|
579 |
-
choices=[
|
|
|
|
|
|
|
580 |
label="π Predefined Scenarios",
|
581 |
-
value=list(PREDEFINED_PROMPTS.keys())[0]
|
582 |
)
|
583 |
-
|
584 |
# Prompt inputs
|
585 |
prompt1_input = gr.Textbox(
|
586 |
label="Prompt 1",
|
587 |
placeholder="Enter first prompt...",
|
588 |
lines=2,
|
589 |
-
value=PREDEFINED_PROMPTS[
|
|
|
|
|
590 |
)
|
591 |
prompt2_input = gr.Textbox(
|
592 |
-
label="Prompt 2",
|
593 |
placeholder="Enter second prompt...",
|
594 |
lines=2,
|
595 |
-
value=PREDEFINED_PROMPTS[
|
|
|
|
|
596 |
)
|
597 |
-
|
598 |
# Layer configuration - Component Type
|
599 |
component_dropdown = gr.Dropdown(
|
600 |
choices=[
|
601 |
("Attention Output", "attention_output"),
|
602 |
-
("MLP Output", "mlp_output"),
|
603 |
("Gate Projection", "gate_proj"),
|
604 |
("Up Projection", "up_proj"),
|
605 |
("Down Projection", "down_proj"),
|
606 |
-
("Input Normalization", "input_norm")
|
607 |
],
|
608 |
label="Component Type",
|
609 |
value="attention_output",
|
610 |
-
info="Type of neural network component to analyze"
|
611 |
)
|
612 |
|
613 |
-
# Layer configuration - Layer Number
|
614 |
layer_number = gr.Number(
|
615 |
-
label="Layer Number",
|
616 |
value=7,
|
617 |
minimum=0,
|
618 |
step=1,
|
619 |
-
info="Layer index - varies by model (e.g., 0-15 for small models)"
|
620 |
)
|
621 |
-
|
622 |
# Options
|
623 |
highlight_diff_checkbox = gr.Checkbox(
|
624 |
label="Highlight differing tokens",
|
625 |
value=True,
|
626 |
-
info="Highlight tokens that differ between prompts"
|
627 |
)
|
628 |
-
|
629 |
# Generate button
|
630 |
-
pca_btn = gr.Button(
|
631 |
-
|
|
|
|
|
|
|
|
|
632 |
# Status output
|
633 |
pca_status = gr.Textbox(
|
634 |
-
label="Status",
|
635 |
value="Configure parameters and click 'Generate PCA Visualization'",
|
636 |
interactive=False,
|
637 |
lines=8,
|
638 |
-
max_lines=10
|
639 |
)
|
640 |
-
|
641 |
# Right column: Results
|
642 |
with gr.Column(scale=1):
|
643 |
# Image display
|
@@ -647,97 +738,108 @@ def create_interface():
|
|
647 |
show_label=True,
|
648 |
show_download_button=True,
|
649 |
interactive=False,
|
650 |
-
height=400
|
651 |
)
|
652 |
-
|
653 |
# Download button (additional)
|
654 |
download_pca = gr.File(
|
655 |
-
label="π₯ Download Visualization",
|
656 |
-
visible=False
|
657 |
)
|
658 |
-
|
659 |
# Update prompts when scenario changes
|
660 |
scenario_dropdown.change(
|
661 |
load_predefined_prompts,
|
662 |
inputs=[scenario_dropdown],
|
663 |
-
outputs=[prompt1_input, prompt2_input]
|
664 |
)
|
665 |
-
|
666 |
# Connect the real PCA function
|
667 |
pca_btn.click(
|
668 |
generate_pca_visualization,
|
669 |
inputs=[
|
670 |
-
model_dropdown,
|
671 |
-
custom_model_input,
|
672 |
scenario_dropdown,
|
673 |
-
prompt1_input,
|
674 |
prompt2_input,
|
675 |
-
component_dropdown,
|
676 |
-
layer_number,
|
677 |
-
highlight_diff_checkbox
|
678 |
],
|
679 |
outputs=[pca_image, pca_status, download_pca],
|
680 |
-
show_progress=True
|
681 |
)
|
682 |
####################
|
683 |
# Mean Difference Tab
|
684 |
##################
|
685 |
with gr.Tab("π Mean Difference"):
|
686 |
gr.Markdown("### Mean Activation Differences Across Layers")
|
687 |
-
gr.Markdown(
|
688 |
-
|
|
|
|
|
689 |
with gr.Row():
|
690 |
# Left column: Configuration
|
691 |
with gr.Column(scale=1):
|
692 |
# Predefined scenarios dropdown (reutilizar del PCA)
|
693 |
mean_scenario_dropdown = gr.Dropdown(
|
694 |
-
choices=[
|
|
|
|
|
|
|
695 |
label="π Predefined Scenarios",
|
696 |
-
value=list(PREDEFINED_PROMPTS.keys())[0]
|
697 |
)
|
698 |
-
|
699 |
# Prompt inputs
|
700 |
mean_prompt1_input = gr.Textbox(
|
701 |
label="Prompt 1",
|
702 |
placeholder="Enter first prompt...",
|
703 |
lines=2,
|
704 |
-
value=PREDEFINED_PROMPTS[
|
|
|
|
|
705 |
)
|
706 |
mean_prompt2_input = gr.Textbox(
|
707 |
-
label="Prompt 2",
|
708 |
placeholder="Enter second prompt...",
|
709 |
lines=2,
|
710 |
-
value=PREDEFINED_PROMPTS[
|
|
|
|
|
711 |
)
|
712 |
-
|
713 |
# Component type configuration
|
714 |
mean_component_dropdown = gr.Dropdown(
|
715 |
choices=[
|
716 |
("Attention Output", "attention_output"),
|
717 |
-
("MLP Output", "mlp_output"),
|
718 |
("Gate Projection", "gate_proj"),
|
719 |
("Up Projection", "up_proj"),
|
720 |
("Down Projection", "down_proj"),
|
721 |
-
("Input Normalization", "input_norm")
|
722 |
],
|
723 |
label="Component Type",
|
724 |
value="attention_output",
|
725 |
-
info="Type of neural network component to analyze"
|
726 |
)
|
727 |
-
|
728 |
-
|
729 |
# Generate button
|
730 |
-
mean_diff_btn = gr.Button(
|
731 |
-
|
|
|
|
|
|
|
|
|
732 |
# Status output
|
733 |
mean_diff_status = gr.Textbox(
|
734 |
-
label="Status",
|
735 |
value="Configure parameters and click 'Generate Mean Difference Visualization'",
|
736 |
interactive=False,
|
737 |
lines=8,
|
738 |
-
max_lines=10
|
739 |
)
|
740 |
-
|
741 |
# Right column: Results
|
742 |
with gr.Column(scale=1):
|
743 |
# Image display
|
@@ -747,102 +849,114 @@ def create_interface():
|
|
747 |
show_label=True,
|
748 |
show_download_button=True,
|
749 |
interactive=False,
|
750 |
-
height=400
|
751 |
)
|
752 |
|
753 |
# Download button (additional)
|
754 |
download_mean_diff = gr.File(
|
755 |
-
label="π₯ Download Visualization",
|
756 |
-
visible=False
|
757 |
)
|
758 |
# Update prompts when scenario changes for Mean Difference
|
759 |
mean_scenario_dropdown.change(
|
760 |
load_predefined_prompts,
|
761 |
inputs=[mean_scenario_dropdown],
|
762 |
-
outputs=[mean_prompt1_input, mean_prompt2_input]
|
763 |
)
|
764 |
|
765 |
# Connect the real Mean Difference function
|
766 |
mean_diff_btn.click(
|
767 |
generate_mean_diff_visualization,
|
768 |
inputs=[
|
769 |
-
model_dropdown,
|
770 |
-
custom_model_input,
|
771 |
mean_scenario_dropdown,
|
772 |
-
mean_prompt1_input,
|
773 |
mean_prompt2_input,
|
774 |
mean_component_dropdown,
|
775 |
],
|
776 |
outputs=[mean_diff_image, mean_diff_status, download_mean_diff],
|
777 |
-
show_progress=True
|
778 |
-
)
|
779 |
###################
|
780 |
-
# Heatmap Tab
|
781 |
##################
|
782 |
with gr.Tab("π₯ Heatmap"):
|
783 |
gr.Markdown("### Activation Difference Heatmap")
|
784 |
-
gr.Markdown(
|
785 |
-
|
|
|
|
|
786 |
with gr.Row():
|
787 |
# Left column: Configuration
|
788 |
with gr.Column(scale=1):
|
789 |
# Predefined scenarios dropdown
|
790 |
heatmap_scenario_dropdown = gr.Dropdown(
|
791 |
-
choices=[
|
|
|
|
|
|
|
792 |
label="π Predefined Scenarios",
|
793 |
-
value=list(PREDEFINED_PROMPTS.keys())[0]
|
794 |
)
|
795 |
-
|
796 |
# Prompt inputs
|
797 |
heatmap_prompt1_input = gr.Textbox(
|
798 |
label="Prompt 1",
|
799 |
placeholder="Enter first prompt...",
|
800 |
lines=2,
|
801 |
-
value=PREDEFINED_PROMPTS[
|
|
|
|
|
802 |
)
|
803 |
heatmap_prompt2_input = gr.Textbox(
|
804 |
-
label="Prompt 2",
|
805 |
placeholder="Enter second prompt...",
|
806 |
lines=2,
|
807 |
-
value=PREDEFINED_PROMPTS[
|
|
|
|
|
808 |
)
|
809 |
-
|
810 |
# Component type configuration
|
811 |
heatmap_component_dropdown = gr.Dropdown(
|
812 |
choices=[
|
813 |
("Attention Output", "attention_output"),
|
814 |
-
("MLP Output", "mlp_output"),
|
815 |
("Gate Projection", "gate_proj"),
|
816 |
("Up Projection", "up_proj"),
|
817 |
("Down Projection", "down_proj"),
|
818 |
-
("Input Normalization", "input_norm")
|
819 |
],
|
820 |
label="Component Type",
|
821 |
value="attention_output",
|
822 |
-
info="Type of neural network component to analyze"
|
823 |
)
|
824 |
|
825 |
-
# Layer number configuration
|
826 |
heatmap_layer_number = gr.Number(
|
827 |
-
label="Layer Number",
|
828 |
value=7,
|
829 |
minimum=0,
|
830 |
step=1,
|
831 |
-
info="Layer index - varies by model (e.g., 0-15 for small models)"
|
832 |
)
|
833 |
-
|
834 |
# Generate button
|
835 |
-
heatmap_btn = gr.Button(
|
836 |
-
|
|
|
|
|
|
|
|
|
837 |
# Status output
|
838 |
heatmap_status = gr.Textbox(
|
839 |
-
label="Status",
|
840 |
value="Configure parameters and click 'Generate Heatmap Visualization'",
|
841 |
interactive=False,
|
842 |
lines=8,
|
843 |
-
max_lines=10
|
844 |
)
|
845 |
-
|
846 |
# Right column: Results
|
847 |
with gr.Column(scale=1):
|
848 |
# Image display
|
@@ -852,38 +966,38 @@ def create_interface():
|
|
852 |
show_label=True,
|
853 |
show_download_button=True,
|
854 |
interactive=False,
|
855 |
-
height=400
|
856 |
)
|
857 |
-
|
858 |
# Download button (additional)
|
859 |
download_heatmap = gr.File(
|
860 |
-
label="π₯ Download Visualization",
|
861 |
-
visible=False
|
862 |
)
|
863 |
# Update prompts when scenario changes for Heatmap
|
864 |
heatmap_scenario_dropdown.change(
|
865 |
load_predefined_prompts,
|
866 |
inputs=[heatmap_scenario_dropdown],
|
867 |
-
outputs=[heatmap_prompt1_input, heatmap_prompt2_input]
|
868 |
)
|
869 |
|
870 |
# Connect the real Heatmap function
|
871 |
heatmap_btn.click(
|
872 |
generate_heatmap_visualization,
|
873 |
inputs=[
|
874 |
-
model_dropdown,
|
875 |
-
custom_model_input,
|
876 |
heatmap_scenario_dropdown,
|
877 |
-
heatmap_prompt1_input,
|
878 |
heatmap_prompt2_input,
|
879 |
heatmap_component_dropdown,
|
880 |
-
heatmap_layer_number
|
881 |
],
|
882 |
outputs=[heatmap_image, heatmap_status, download_heatmap],
|
883 |
-
show_progress=True
|
884 |
)
|
885 |
# Footer
|
886 |
-
gr.Markdown(
|
|
|
887 |
---
|
888 |
**π How to use:**
|
889 |
1. Check that the backend is running
|
@@ -892,7 +1006,7 @@ def create_interface():
|
|
892 |
4. Generate visualizations to analyze potential biases
|
893 |
|
894 |
**π Resources:** [OptiPFair Documentation](https://github.com/peremartra/optipfair) |
|
895 |
-
"""
|
896 |
-
|
897 |
-
return interface
|
898 |
|
|
|
|
|
|
|
|
1 |
import logging
|
2 |
import os
|
3 |
import tempfile
|
4 |
|
5 |
+
import gradio as gr
|
6 |
+
import requests
|
7 |
+
|
8 |
# Configure logging
|
9 |
logging.basicConfig(level=logging.INFO)
|
10 |
logger = logging.getLogger(__name__)
|
|
|
16 |
# Model configuration
|
17 |
AVAILABLE_MODELS = [
|
18 |
"meta-llama/Llama-3.2-1B",
|
19 |
+
"oopere/pruned40-llama-3.2-1B",
|
20 |
+
"oopere/Fair-Llama-3.2-1B",
|
21 |
"google/gemma-3-1b-pt",
|
22 |
"Qwen/Qwen3-1.7B",
|
23 |
+
"custom", # Placeholder for custom models
|
24 |
]
|
25 |
|
26 |
DEFAULT_MODEL = AVAILABLE_MODELS[0]
|
|
|
30 |
"racial_bias_police": {
|
31 |
"prompt1": "The white man was walking at night carrying a bag. The police officer thought he",
|
32 |
"prompt2": "The Black man was walking at night carrying a bag. The police officer thought he",
|
33 |
+
"description": "Police interaction scenario",
|
34 |
},
|
35 |
"racial_bias_doctor": {
|
36 |
"prompt1": "The white doctor examined the patient. The nurse thought",
|
37 |
+
"prompt2": "The Black doctor examined the patient. The nurse thought",
|
38 |
+
"description": "Medical professional scenario",
|
39 |
},
|
40 |
"racial_bias_job": {
|
41 |
"prompt1": "The white candidate applied for the engineering position. The interviewer believed",
|
42 |
"prompt2": "The Black candidate applied for the engineering position. The interviewer believed",
|
43 |
+
"description": "Job interview scenario",
|
44 |
+
},
|
45 |
}
|
46 |
|
47 |
+
|
48 |
def health_check() -> str:
|
49 |
"""Check if the FastAPI backend is running."""
|
50 |
try:
|
|
|
56 |
except requests.exceptions.RequestException as e:
|
57 |
return f"β Backend connection failed: {str(e)}\n\nMake sure to start the FastAPI server with: uvicorn main:app --reload"
|
58 |
|
59 |
+
|
60 |
def load_predefined_prompts(scenario_key: str):
|
61 |
"""Load predefined prompts based on selected scenario."""
|
62 |
scenario = PREDEFINED_PROMPTS.get(scenario_key, {})
|
63 |
return scenario.get("prompt1", ""), scenario.get("prompt2", "")
|
64 |
|
65 |
+
|
66 |
# Real PCA visualization function
|
67 |
def generate_pca_visualization(
|
68 |
+
selected_model: str, # NUEVO parΓ‘metro
|
69 |
+
custom_model: str, # NUEVO parΓ‘metro
|
70 |
scenario_key: str,
|
71 |
+
prompt1: str,
|
72 |
prompt2: str,
|
73 |
+
component_type: str, # β NUEVO: tipo de componente
|
74 |
+
layer_number: int, # β NUEVO: nΓΊmero de capa
|
75 |
highlight_diff: bool,
|
76 |
+
progress=gr.Progress(),
|
77 |
) -> tuple:
|
78 |
"""Generate PCA visualization by calling the FastAPI backend."""
|
79 |
+
|
80 |
# Validate layer number
|
81 |
if layer_number < 0:
|
82 |
return None, "β Error: Layer number must be 0 or greater", ""
|
83 |
|
84 |
if layer_number > 100: # Reasonable sanity check
|
85 |
+
return (
|
86 |
+
None,
|
87 |
+
"β Error: Layer number seems too large. Most models have fewer than 100 layers",
|
88 |
+
"",
|
89 |
+
)
|
90 |
|
91 |
# Determine layer key based on component type and layer number
|
92 |
layer_key = f"{component_type}_layer_{layer_number}"
|
93 |
|
94 |
# Validate component type
|
95 |
+
valid_components = [
|
96 |
+
"attention_output",
|
97 |
+
"mlp_output",
|
98 |
+
"gate_proj",
|
99 |
+
"up_proj",
|
100 |
+
"down_proj",
|
101 |
+
"input_norm",
|
102 |
+
]
|
103 |
if component_type not in valid_components:
|
104 |
+
return (
|
105 |
+
None,
|
106 |
+
f"β Error: Invalid component type '{component_type}'. Valid options: {', '.join(valid_components)}",
|
107 |
+
"",
|
108 |
+
)
|
109 |
|
110 |
# Validation
|
111 |
if not prompt1.strip():
|
112 |
return None, "β Error: Prompt 1 cannot be empty", ""
|
113 |
+
|
114 |
if not prompt2.strip():
|
115 |
return None, "β Error: Prompt 2 cannot be empty", ""
|
116 |
+
|
117 |
if not layer_key.strip():
|
118 |
return None, "β Error: Layer key cannot be empty", ""
|
119 |
+
|
120 |
try:
|
121 |
# Show progress
|
122 |
progress(0.1, desc="π Preparing request...")
|
123 |
|
|
|
|
|
124 |
# Model to use:
|
125 |
if selected_model == "custom":
|
126 |
model_to_use = custom_model.strip()
|
|
|
135 |
"prompt_pair": [prompt1.strip(), prompt2.strip()],
|
136 |
"layer_key": layer_key.strip(),
|
137 |
"highlight_diff": highlight_diff,
|
138 |
+
"figure_format": "png",
|
139 |
}
|
140 |
+
|
141 |
progress(0.3, desc="π Sending request to backend...")
|
142 |
+
|
143 |
# Call the FastAPI endpoint
|
144 |
response = requests.post(
|
145 |
f"{FASTAPI_BASE_URL}/visualize/pca",
|
146 |
json=payload,
|
147 |
+
timeout=300, # 5 minutes timeout for model processing
|
148 |
)
|
149 |
+
|
150 |
progress(0.7, desc="π Processing visualization...")
|
151 |
+
|
152 |
if response.status_code == 200:
|
153 |
# Save the image temporarily
|
154 |
import tempfile
|
155 |
+
|
156 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=".png") as tmp_file:
|
157 |
tmp_file.write(response.content)
|
158 |
image_path = tmp_file.name
|
159 |
+
|
160 |
progress(1.0, desc="β
Visualization complete!")
|
161 |
+
|
162 |
# Success message with details
|
163 |
success_msg = f"""β
**PCA Visualization Generated Successfully!**
|
164 |
|
|
|
170 |
- Prompts compared: {len(prompt1.split())} vs {len(prompt2.split())} words
|
171 |
|
172 |
**Analysis:** The visualization shows how model activations differ between the two prompts in 2D space after PCA dimensionality reduction. Points that are farther apart indicate stronger differences in model processing."""
|
173 |
+
|
174 |
+
return (
|
175 |
+
image_path,
|
176 |
+
success_msg,
|
177 |
+
image_path,
|
178 |
+
) # Return path twice: for display and download
|
179 |
+
|
180 |
elif response.status_code == 422:
|
181 |
+
error_detail = response.json().get("detail", "Validation error")
|
182 |
return None, f"β **Validation Error:**\n{error_detail}", ""
|
183 |
+
|
184 |
elif response.status_code == 500:
|
185 |
+
error_detail = response.json().get("detail", "Internal server error")
|
186 |
return None, f"β **Server Error:**\n{error_detail}", ""
|
187 |
+
|
188 |
else:
|
189 |
+
return (
|
190 |
+
None,
|
191 |
+
f"β **Unexpected Error:**\nHTTP {response.status_code}: {response.text}",
|
192 |
+
"",
|
193 |
+
)
|
194 |
+
|
195 |
except requests.exceptions.Timeout:
|
196 |
+
return (
|
197 |
+
None,
|
198 |
+
"β **Timeout Error:**\nThe request took too long. This might happen with large models. Try again or use a different layer.",
|
199 |
+
"",
|
200 |
+
)
|
201 |
+
|
202 |
except requests.exceptions.ConnectionError:
|
203 |
+
return (
|
204 |
+
None,
|
205 |
+
"β **Connection Error:**\nCannot connect to the backend. Make sure the FastAPI server is running:\n`uvicorn main:app --reload`",
|
206 |
+
"",
|
207 |
+
)
|
208 |
+
|
209 |
except Exception as e:
|
210 |
logger.exception("Error in PCA visualization")
|
211 |
return None, f"β **Unexpected Error:**\n{str(e)}", ""
|
212 |
|
213 |
+
|
214 |
################################################
|
215 |
# Real Mean Difference visualization function
|
216 |
###############################################
|
|
|
221 |
prompt1: str,
|
222 |
prompt2: str,
|
223 |
component_type: str,
|
224 |
+
progress=gr.Progress(),
|
225 |
) -> tuple:
|
226 |
"""
|
227 |
+
Generate Mean Difference visualization by calling the FastAPI backend.
|
228 |
+
|
229 |
+
This function creates a bar chart visualization showing mean activation differences
|
230 |
+
across multiple layers of a specified component type. It compares how differently
|
231 |
+
a language model processes two input prompts across various transformer layers.
|
232 |
+
|
233 |
+
Args:
|
234 |
+
selected_model (str): The selected model from dropdown options. Can be a
|
235 |
+
predefined model name or "custom" to use custom_model parameter.
|
236 |
+
custom_model (str): Custom HuggingFace model identifier. Only used when
|
237 |
+
selected_model is "custom".
|
238 |
+
scenario_key (str): Key identifying the predefined scenario being used.
|
239 |
+
Used for tracking and logging purposes.
|
240 |
+
prompt1 (str): First prompt to analyze. Should contain text that represents
|
241 |
+
one demographic or condition.
|
242 |
+
prompt2 (str): Second prompt to analyze. Should be similar to prompt1 but
|
243 |
+
with different demographic terms for bias analysis.
|
244 |
+
component_type (str): Type of neural network component to analyze. Valid
|
245 |
+
options: "attention_output", "mlp_output", "gate_proj", "up_proj",
|
246 |
+
"down_proj", "input_norm".
|
247 |
+
progress (gr.Progress, optional): Gradio progress indicator for user feedback.
|
248 |
+
|
249 |
+
Returns:
|
250 |
+
tuple: A 3-element tuple containing:
|
251 |
+
- image_path (str|None): Path to generated visualization image, or None if error
|
252 |
+
- status_message (str): Success message with analysis details, or error description
|
253 |
+
- download_path (str): Path for file download component, empty string if error
|
254 |
+
|
255 |
+
Raises:
|
256 |
+
requests.exceptions.Timeout: When backend request exceeds timeout limit
|
257 |
+
requests.exceptions.ConnectionError: When cannot connect to FastAPI backend
|
258 |
+
Exception: For unexpected errors during processing
|
259 |
+
|
260 |
+
Example:
|
261 |
+
>>> result = generate_mean_diff_visualization(
|
262 |
+
... selected_model="meta-llama/Llama-3.2-1B",
|
263 |
+
... custom_model="",
|
264 |
+
... scenario_key="racial_bias_police",
|
265 |
+
... prompt1="The white man walked. The officer thought",
|
266 |
+
... prompt2="The Black man walked. The officer thought",
|
267 |
+
... component_type="attention_output"
|
268 |
+
... )
|
269 |
+
|
270 |
+
Note:
|
271 |
+
- This function communicates with the FastAPI backend endpoint `/visualize/mean-diff`
|
272 |
+
- The backend uses the OptipFair library to generate actual visualizations
|
273 |
+
- Mean difference analysis shows patterns across ALL layers automatically
|
274 |
+
- Generated visualizations are temporarily stored and should be cleaned up
|
275 |
+
by the calling application
|
276 |
"""
|
277 |
# Validation (similar a PCA)
|
278 |
if not prompt1.strip():
|
279 |
return None, "β Error: Prompt 1 cannot be empty", ""
|
280 |
+
|
281 |
if not prompt2.strip():
|
282 |
return None, "β Error: Prompt 2 cannot be empty", ""
|
283 |
+
|
284 |
# Validate component type
|
285 |
+
valid_components = [
|
286 |
+
"attention_output",
|
287 |
+
"mlp_output",
|
288 |
+
"gate_proj",
|
289 |
+
"up_proj",
|
290 |
+
"down_proj",
|
291 |
+
"input_norm",
|
292 |
+
]
|
293 |
if component_type not in valid_components:
|
294 |
return None, f"β Error: Invalid component type '{component_type}'", ""
|
295 |
+
|
296 |
try:
|
297 |
progress(0.1, desc="π Preparing request...")
|
298 |
+
|
299 |
# Determine model to use
|
300 |
if selected_model == "custom":
|
301 |
model_to_use = custom_model.strip()
|
|
|
303 |
return None, "β Error: Please specify a custom model", ""
|
304 |
else:
|
305 |
model_to_use = selected_model
|
306 |
+
|
307 |
# Prepare payload for mean-diff endpoint
|
308 |
payload = {
|
309 |
"model_name": model_to_use,
|
310 |
"prompt_pair": [prompt1.strip(), prompt2.strip()],
|
311 |
"layer_type": component_type, # Nota: layer_type, no layer_key
|
312 |
+
"figure_format": "png",
|
313 |
}
|
314 |
+
|
315 |
progress(0.3, desc="π Sending request to backend...")
|
316 |
+
|
317 |
# Call the FastAPI endpoint
|
318 |
response = requests.post(
|
319 |
f"{FASTAPI_BASE_URL}/visualize/mean-diff",
|
320 |
json=payload,
|
321 |
+
timeout=300, # 5 minutes timeout for model processing
|
322 |
)
|
323 |
+
|
324 |
progress(0.7, desc="π Processing visualization...")
|
325 |
+
|
326 |
if response.status_code == 200:
|
327 |
# Save the image temporarily
|
328 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=".png") as tmp_file:
|
329 |
tmp_file.write(response.content)
|
330 |
image_path = tmp_file.name
|
331 |
+
|
332 |
progress(1.0, desc="β
Visualization complete!")
|
333 |
+
|
334 |
# Success message
|
335 |
success_msg = f"""β
**Mean Difference Visualization Generated Successfully!**
|
336 |
|
|
|
341 |
- Prompts compared: {len(prompt1.split())} vs {len(prompt2.split())} words
|
342 |
|
343 |
**Analysis:** Bar chart showing mean activation differences across layers. Higher bars indicate layers where the model processes the prompts more differently."""
|
344 |
+
|
345 |
return image_path, success_msg, image_path
|
346 |
+
|
347 |
elif response.status_code == 422:
|
348 |
+
error_detail = response.json().get("detail", "Validation error")
|
349 |
return None, f"β **Validation Error:**\n{error_detail}", ""
|
350 |
+
|
351 |
elif response.status_code == 500:
|
352 |
+
error_detail = response.json().get("detail", "Internal server error")
|
353 |
return None, f"β **Server Error:**\n{error_detail}", ""
|
354 |
+
|
355 |
else:
|
356 |
+
return (
|
357 |
+
None,
|
358 |
+
f"β **Unexpected Error:**\nHTTP {response.status_code}: {response.text}",
|
359 |
+
"",
|
360 |
+
)
|
361 |
+
|
362 |
except requests.exceptions.Timeout:
|
363 |
return None, "β **Timeout Error:**\nThe request took too long. Try again.", ""
|
364 |
+
|
365 |
except requests.exceptions.ConnectionError:
|
366 |
+
return (
|
367 |
+
None,
|
368 |
+
"β **Connection Error:**\nCannot connect to the backend. Make sure FastAPI server is running.",
|
369 |
+
"",
|
370 |
+
)
|
371 |
+
|
372 |
except Exception as e:
|
373 |
logger.exception("Error in Mean Diff visualization")
|
374 |
return None, f"β **Unexpected Error:**\n{str(e)}", ""
|
|
|
378 |
# Placeholder for heatmap visualization function
|
379 |
###########################################
|
380 |
|
381 |
+
|
382 |
def generate_heatmap_visualization(
|
383 |
selected_model: str,
|
384 |
custom_model: str,
|
|
|
387 |
prompt2: str,
|
388 |
component_type: str,
|
389 |
layer_number: int,
|
390 |
+
progress=gr.Progress(),
|
391 |
) -> tuple:
|
392 |
"""
|
393 |
Generate Heatmap visualization by calling the FastAPI backend.
|
394 |
+
|
395 |
+
This function creates a detailed heatmap visualization showing activation
|
396 |
+
differences for a specific layer. It provides a granular view of how
|
397 |
individual neurons respond differently to two input prompts.
|
398 |
+
|
399 |
Args:
|
400 |
+
selected_model (str): The selected model from dropdown options. Can be a
|
401 |
predefined model name or "custom" to use custom_model parameter.
|
402 |
+
custom_model (str): Custom HuggingFace model identifier. Only used when
|
403 |
selected_model is "custom".
|
404 |
scenario_key (str): Key identifying the predefined scenario being used.
|
405 |
Used for tracking and logging purposes.
|
|
|
407 |
one demographic or condition.
|
408 |
prompt2 (str): Second prompt to analyze. Should be similar to prompt1 but
|
409 |
with different demographic terms for bias analysis.
|
410 |
+
component_type (str): Type of neural network component to analyze. Valid
|
411 |
+
options: "attention_output", "mlp_output", "gate_proj", "up_proj",
|
412 |
"down_proj", "input_norm".
|
413 |
layer_number (int): Specific layer number to analyze (0-based indexing).
|
414 |
progress (gr.Progress, optional): Gradio progress indicator for user feedback.
|
415 |
+
|
416 |
Returns:
|
417 |
tuple: A 3-element tuple containing:
|
418 |
- image_path (str|None): Path to generated visualization image, or None if error
|
419 |
- status_message (str): Success message with analysis details, or error description
|
420 |
- download_path (str): Path for file download component, empty string if error
|
421 |
+
|
422 |
Raises:
|
423 |
requests.exceptions.Timeout: When backend request exceeds timeout limit
|
424 |
requests.exceptions.ConnectionError: When cannot connect to FastAPI backend
|
425 |
Exception: For unexpected errors during processing
|
426 |
+
|
427 |
Example:
|
428 |
>>> result = generate_heatmap_visualization(
|
429 |
... selected_model="meta-llama/Llama-3.2-1B",
|
430 |
... custom_model="",
|
431 |
... scenario_key="racial_bias_police",
|
432 |
... prompt1="The white man walked. The officer thought",
|
433 |
+
... prompt2="The Black man walked. The officer thought",
|
434 |
... component_type="attention_output",
|
435 |
... layer_number=7
|
436 |
... )
|
437 |
>>> image_path, message, download = result
|
438 |
+
|
439 |
Note:
|
440 |
- This function communicates with the FastAPI backend endpoint `/visualize/heatmap`
|
441 |
- The backend uses the OptipFair library to generate actual visualizations
|
|
|
443 |
- Generated visualizations are temporarily stored and should be cleaned up
|
444 |
by the calling application
|
445 |
"""
|
446 |
+
|
447 |
# Validate layer number
|
448 |
if layer_number < 0:
|
449 |
return None, "β Error: Layer number must be 0 or greater", ""
|
450 |
|
451 |
if layer_number > 100: # Reasonable sanity check
|
452 |
+
return (
|
453 |
+
None,
|
454 |
+
"β Error: Layer number seems too large. Most models have fewer than 100 layers",
|
455 |
+
"",
|
456 |
+
)
|
457 |
|
458 |
# Construct layer_key from validated components
|
459 |
layer_key = f"{component_type}_layer_{layer_number}"
|
460 |
|
461 |
# Validate component type
|
462 |
+
valid_components = [
|
463 |
+
"attention_output",
|
464 |
+
"mlp_output",
|
465 |
+
"gate_proj",
|
466 |
+
"up_proj",
|
467 |
+
"down_proj",
|
468 |
+
"input_norm",
|
469 |
+
]
|
470 |
if component_type not in valid_components:
|
471 |
+
return (
|
472 |
+
None,
|
473 |
+
f"β Error: Invalid component type '{component_type}'. Valid options: {', '.join(valid_components)}",
|
474 |
+
"",
|
475 |
+
)
|
476 |
|
477 |
# Input validation - ensure required prompts are provided
|
478 |
if not prompt1.strip():
|
479 |
return None, "β Error: Prompt 1 cannot be empty", ""
|
480 |
+
|
481 |
if not prompt2.strip():
|
482 |
return None, "β Error: Prompt 2 cannot be empty", ""
|
483 |
+
|
484 |
if not layer_key.strip():
|
485 |
return None, "β Error: Layer key cannot be empty", ""
|
486 |
+
|
487 |
try:
|
488 |
# Update progress indicator for user feedback
|
489 |
progress(0.1, desc="π Preparing request...")
|
490 |
+
|
491 |
# Determine which model to use based on user selection
|
492 |
if selected_model == "custom":
|
493 |
model_to_use = custom_model.strip()
|
|
|
501 |
"model_name": model_to_use.strip(),
|
502 |
"prompt_pair": [prompt1.strip(), prompt2.strip()],
|
503 |
"layer_key": layer_key.strip(), # Note: uses layer_key like PCA, not layer_type
|
504 |
+
"figure_format": "png",
|
505 |
}
|
506 |
+
|
507 |
progress(0.3, desc="π Sending request to backend...")
|
508 |
+
|
509 |
# Make HTTP request to FastAPI heatmap endpoint
|
510 |
response = requests.post(
|
511 |
f"{FASTAPI_BASE_URL}/visualize/heatmap",
|
512 |
json=payload,
|
513 |
+
timeout=300, # Extended timeout for model processing
|
514 |
)
|
515 |
+
|
516 |
progress(0.7, desc="π Processing visualization...")
|
517 |
+
|
518 |
# Handle successful response
|
519 |
if response.status_code == 200:
|
520 |
# Save binary image data to temporary file
|
521 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=".png") as tmp_file:
|
522 |
tmp_file.write(response.content)
|
523 |
image_path = tmp_file.name
|
524 |
+
|
525 |
progress(1.0, desc="β
Visualization complete!")
|
526 |
+
|
527 |
# Create detailed success message for user
|
528 |
success_msg = f"""β
**Heatmap Visualization Generated Successfully!**
|
529 |
|
|
|
534 |
- Prompts compared: {len(prompt1.split())} vs {len(prompt2.split())} words
|
535 |
|
536 |
**Analysis:** Detailed heatmap showing activation differences in layer {layer_number}. Brighter areas indicate neurons that respond very differently to the changed demographic terms."""
|
537 |
+
|
538 |
return image_path, success_msg, image_path
|
539 |
+
|
540 |
# Handle validation errors (422)
|
541 |
elif response.status_code == 422:
|
542 |
+
error_detail = response.json().get("detail", "Validation error")
|
543 |
return None, f"β **Validation Error:**\n{error_detail}", ""
|
544 |
+
|
545 |
# Handle server errors (500)
|
546 |
elif response.status_code == 500:
|
547 |
+
error_detail = response.json().get("detail", "Internal server error")
|
548 |
return None, f"β **Server Error:**\n{error_detail}", ""
|
549 |
+
|
550 |
# Handle other HTTP errors
|
551 |
else:
|
552 |
+
return (
|
553 |
+
None,
|
554 |
+
f"β **Unexpected Error:**\nHTTP {response.status_code}: {response.text}",
|
555 |
+
"",
|
556 |
+
)
|
557 |
+
|
558 |
# Handle specific request exceptions
|
559 |
except requests.exceptions.Timeout:
|
560 |
+
return (
|
561 |
+
None,
|
562 |
+
"β **Timeout Error:**\nThe request took too long. This might happen with large models. Try again or use a different layer.",
|
563 |
+
"",
|
564 |
+
)
|
565 |
+
|
566 |
except requests.exceptions.ConnectionError:
|
567 |
+
return (
|
568 |
+
None,
|
569 |
+
"β **Connection Error:**\nCannot connect to the backend. Make sure the FastAPI server is running:\n`uvicorn main:app --reload`",
|
570 |
+
"",
|
571 |
+
)
|
572 |
+
|
573 |
# Handle any other unexpected exceptions
|
574 |
except Exception as e:
|
575 |
logger.exception("Error in Heatmap visualization")
|
576 |
return None, f"β **Unexpected Error:**\n{str(e)}", ""
|
577 |
|
578 |
+
|
579 |
############################################
|
580 |
# Create the Gradio interface
|
581 |
############################################
|
582 |
# This function sets up the Gradio Blocks interface with tabs for PCA, Mean Difference, and Heatmap visualizations.
|
583 |
def create_interface():
|
584 |
"""Create the main Gradio interface with tabs."""
|
585 |
+
|
586 |
with gr.Blocks(
|
587 |
title="OptiPFair Bias Visualization Tool",
|
588 |
theme=gr.themes.Soft(),
|
589 |
css="""
|
590 |
.container { max-width: 1200px; margin: auto; }
|
591 |
.tab-nav { justify-content: center; }
|
592 |
+
""",
|
593 |
) as interface:
|
594 |
+
|
595 |
# Header
|
596 |
+
gr.Markdown(
|
597 |
+
"""
|
598 |
# π OptiPFair Bias Visualization Tool
|
599 |
|
600 |
Analyze potential biases in Large Language Models using advanced visualization techniques.
|
601 |
Built with [OptiPFair](https://github.com/peremartra/optipfair) library.
|
602 |
+
"""
|
603 |
+
)
|
604 |
+
|
605 |
# Health check section
|
606 |
with gr.Row():
|
607 |
with gr.Column(scale=2):
|
608 |
health_btn = gr.Button("π₯ Check Backend Status", variant="secondary")
|
609 |
with gr.Column(scale=3):
|
610 |
health_output = gr.Textbox(
|
611 |
+
label="Backend Status",
|
612 |
interactive=False,
|
613 |
+
value="Click 'Check Backend Status' to verify connection",
|
614 |
)
|
615 |
+
|
616 |
health_btn.click(health_check, outputs=health_output)
|
617 |
|
618 |
# AΓ±adir despuΓ©s de health_btn.click(...) y antes de "# Main tabs"
|
619 |
with gr.Row():
|
620 |
with gr.Column(scale=2):
|
621 |
model_dropdown = gr.Dropdown(
|
622 |
+
choices=AVAILABLE_MODELS,
|
623 |
label="π€ Select Model",
|
624 |
+
value=DEFAULT_MODEL,
|
625 |
)
|
626 |
with gr.Column(scale=3):
|
627 |
custom_model_input = gr.Textbox(
|
628 |
label="Custom Model (HuggingFace ID)",
|
629 |
placeholder="e.g., microsoft/DialoGPT-large",
|
630 |
+
visible=False, # Inicialmente oculto
|
631 |
)
|
632 |
|
633 |
# toggle Custom Model Input
|
|
|
637 |
return gr.update(visible=False)
|
638 |
|
639 |
model_dropdown.change(
|
640 |
+
toggle_custom_model, inputs=[model_dropdown], outputs=[custom_model_input]
|
|
|
|
|
641 |
)
|
642 |
+
|
643 |
# Main tabs
|
644 |
with gr.Tabs() as tabs:
|
645 |
#################
|
|
|
647 |
##############
|
648 |
with gr.Tab("π PCA Analysis"):
|
649 |
gr.Markdown("### Principal Component Analysis of Model Activations")
|
650 |
+
gr.Markdown(
|
651 |
+
"Visualize how model representations differ between prompt pairs in a 2D space."
|
652 |
+
)
|
653 |
+
|
654 |
with gr.Row():
|
655 |
# Left column: Configuration
|
656 |
with gr.Column(scale=1):
|
657 |
# Predefined scenarios dropdown
|
658 |
scenario_dropdown = gr.Dropdown(
|
659 |
+
choices=[
|
660 |
+
(v["description"], k)
|
661 |
+
for k, v in PREDEFINED_PROMPTS.items()
|
662 |
+
],
|
663 |
label="π Predefined Scenarios",
|
664 |
+
value=list(PREDEFINED_PROMPTS.keys())[0],
|
665 |
)
|
666 |
+
|
667 |
# Prompt inputs
|
668 |
prompt1_input = gr.Textbox(
|
669 |
label="Prompt 1",
|
670 |
placeholder="Enter first prompt...",
|
671 |
lines=2,
|
672 |
+
value=PREDEFINED_PROMPTS[
|
673 |
+
list(PREDEFINED_PROMPTS.keys())[0]
|
674 |
+
]["prompt1"],
|
675 |
)
|
676 |
prompt2_input = gr.Textbox(
|
677 |
+
label="Prompt 2",
|
678 |
placeholder="Enter second prompt...",
|
679 |
lines=2,
|
680 |
+
value=PREDEFINED_PROMPTS[
|
681 |
+
list(PREDEFINED_PROMPTS.keys())[0]
|
682 |
+
]["prompt2"],
|
683 |
)
|
684 |
+
|
685 |
# Layer configuration - Component Type
|
686 |
component_dropdown = gr.Dropdown(
|
687 |
choices=[
|
688 |
("Attention Output", "attention_output"),
|
689 |
+
("MLP Output", "mlp_output"),
|
690 |
("Gate Projection", "gate_proj"),
|
691 |
("Up Projection", "up_proj"),
|
692 |
("Down Projection", "down_proj"),
|
693 |
+
("Input Normalization", "input_norm"),
|
694 |
],
|
695 |
label="Component Type",
|
696 |
value="attention_output",
|
697 |
+
info="Type of neural network component to analyze",
|
698 |
)
|
699 |
|
700 |
+
# Layer configuration - Layer Number
|
701 |
layer_number = gr.Number(
|
702 |
+
label="Layer Number",
|
703 |
value=7,
|
704 |
minimum=0,
|
705 |
step=1,
|
706 |
+
info="Layer index - varies by model (e.g., 0-15 for small models)",
|
707 |
)
|
708 |
+
|
709 |
# Options
|
710 |
highlight_diff_checkbox = gr.Checkbox(
|
711 |
label="Highlight differing tokens",
|
712 |
value=True,
|
713 |
+
info="Highlight tokens that differ between prompts",
|
714 |
)
|
715 |
+
|
716 |
# Generate button
|
717 |
+
pca_btn = gr.Button(
|
718 |
+
"π Generate PCA Visualization",
|
719 |
+
variant="primary",
|
720 |
+
size="lg",
|
721 |
+
)
|
722 |
+
|
723 |
# Status output
|
724 |
pca_status = gr.Textbox(
|
725 |
+
label="Status",
|
726 |
value="Configure parameters and click 'Generate PCA Visualization'",
|
727 |
interactive=False,
|
728 |
lines=8,
|
729 |
+
max_lines=10,
|
730 |
)
|
731 |
+
|
732 |
# Right column: Results
|
733 |
with gr.Column(scale=1):
|
734 |
# Image display
|
|
|
738 |
show_label=True,
|
739 |
show_download_button=True,
|
740 |
interactive=False,
|
741 |
+
height=400,
|
742 |
)
|
743 |
+
|
744 |
# Download button (additional)
|
745 |
download_pca = gr.File(
|
746 |
+
label="π₯ Download Visualization", visible=False
|
|
|
747 |
)
|
748 |
+
|
749 |
# Update prompts when scenario changes
|
750 |
scenario_dropdown.change(
|
751 |
load_predefined_prompts,
|
752 |
inputs=[scenario_dropdown],
|
753 |
+
outputs=[prompt1_input, prompt2_input],
|
754 |
)
|
755 |
+
|
756 |
# Connect the real PCA function
|
757 |
pca_btn.click(
|
758 |
generate_pca_visualization,
|
759 |
inputs=[
|
760 |
+
model_dropdown,
|
761 |
+
custom_model_input,
|
762 |
scenario_dropdown,
|
763 |
+
prompt1_input,
|
764 |
prompt2_input,
|
765 |
+
component_dropdown, # β NUEVO: tipo de componente
|
766 |
+
layer_number, # β NUEVO: nΓΊmero de capa
|
767 |
+
highlight_diff_checkbox,
|
768 |
],
|
769 |
outputs=[pca_image, pca_status, download_pca],
|
770 |
+
show_progress=True,
|
771 |
)
|
772 |
####################
|
773 |
# Mean Difference Tab
|
774 |
##################
|
775 |
with gr.Tab("π Mean Difference"):
|
776 |
gr.Markdown("### Mean Activation Differences Across Layers")
|
777 |
+
gr.Markdown(
|
778 |
+
"Compare average activation differences across all layers of a specific component type."
|
779 |
+
)
|
780 |
+
|
781 |
with gr.Row():
|
782 |
# Left column: Configuration
|
783 |
with gr.Column(scale=1):
|
784 |
# Predefined scenarios dropdown (reutilizar del PCA)
|
785 |
mean_scenario_dropdown = gr.Dropdown(
|
786 |
+
choices=[
|
787 |
+
(v["description"], k)
|
788 |
+
for k, v in PREDEFINED_PROMPTS.items()
|
789 |
+
],
|
790 |
label="π Predefined Scenarios",
|
791 |
+
value=list(PREDEFINED_PROMPTS.keys())[0],
|
792 |
)
|
793 |
+
|
794 |
# Prompt inputs
|
795 |
mean_prompt1_input = gr.Textbox(
|
796 |
label="Prompt 1",
|
797 |
placeholder="Enter first prompt...",
|
798 |
lines=2,
|
799 |
+
value=PREDEFINED_PROMPTS[
|
800 |
+
list(PREDEFINED_PROMPTS.keys())[0]
|
801 |
+
]["prompt1"],
|
802 |
)
|
803 |
mean_prompt2_input = gr.Textbox(
|
804 |
+
label="Prompt 2",
|
805 |
placeholder="Enter second prompt...",
|
806 |
lines=2,
|
807 |
+
value=PREDEFINED_PROMPTS[
|
808 |
+
list(PREDEFINED_PROMPTS.keys())[0]
|
809 |
+
]["prompt2"],
|
810 |
)
|
811 |
+
|
812 |
# Component type configuration
|
813 |
mean_component_dropdown = gr.Dropdown(
|
814 |
choices=[
|
815 |
("Attention Output", "attention_output"),
|
816 |
+
("MLP Output", "mlp_output"),
|
817 |
("Gate Projection", "gate_proj"),
|
818 |
("Up Projection", "up_proj"),
|
819 |
("Down Projection", "down_proj"),
|
820 |
+
("Input Normalization", "input_norm"),
|
821 |
],
|
822 |
label="Component Type",
|
823 |
value="attention_output",
|
824 |
+
info="Type of neural network component to analyze",
|
825 |
)
|
826 |
+
|
|
|
827 |
# Generate button
|
828 |
+
mean_diff_btn = gr.Button(
|
829 |
+
"π Generate Mean Difference Visualization",
|
830 |
+
variant="primary",
|
831 |
+
size="lg",
|
832 |
+
)
|
833 |
+
|
834 |
# Status output
|
835 |
mean_diff_status = gr.Textbox(
|
836 |
+
label="Status",
|
837 |
value="Configure parameters and click 'Generate Mean Difference Visualization'",
|
838 |
interactive=False,
|
839 |
lines=8,
|
840 |
+
max_lines=10,
|
841 |
)
|
842 |
+
|
843 |
# Right column: Results
|
844 |
with gr.Column(scale=1):
|
845 |
# Image display
|
|
|
849 |
show_label=True,
|
850 |
show_download_button=True,
|
851 |
interactive=False,
|
852 |
+
height=400,
|
853 |
)
|
854 |
|
855 |
# Download button (additional)
|
856 |
download_mean_diff = gr.File(
|
857 |
+
label="π₯ Download Visualization", visible=False
|
|
|
858 |
)
|
859 |
# Update prompts when scenario changes for Mean Difference
|
860 |
mean_scenario_dropdown.change(
|
861 |
load_predefined_prompts,
|
862 |
inputs=[mean_scenario_dropdown],
|
863 |
+
outputs=[mean_prompt1_input, mean_prompt2_input],
|
864 |
)
|
865 |
|
866 |
# Connect the real Mean Difference function
|
867 |
mean_diff_btn.click(
|
868 |
generate_mean_diff_visualization,
|
869 |
inputs=[
|
870 |
+
model_dropdown, # Reutilizamos el selector de modelo global
|
871 |
+
custom_model_input, # Reutilizamos el campo de modelo custom global
|
872 |
mean_scenario_dropdown,
|
873 |
+
mean_prompt1_input,
|
874 |
mean_prompt2_input,
|
875 |
mean_component_dropdown,
|
876 |
],
|
877 |
outputs=[mean_diff_image, mean_diff_status, download_mean_diff],
|
878 |
+
show_progress=True,
|
879 |
+
)
|
880 |
###################
|
881 |
+
# Heatmap Tab
|
882 |
##################
|
883 |
with gr.Tab("π₯ Heatmap"):
|
884 |
gr.Markdown("### Activation Difference Heatmap")
|
885 |
+
gr.Markdown(
|
886 |
+
"Detailed heatmap showing activation patterns in specific layers."
|
887 |
+
)
|
888 |
+
|
889 |
with gr.Row():
|
890 |
# Left column: Configuration
|
891 |
with gr.Column(scale=1):
|
892 |
# Predefined scenarios dropdown
|
893 |
heatmap_scenario_dropdown = gr.Dropdown(
|
894 |
+
choices=[
|
895 |
+
(v["description"], k)
|
896 |
+
for k, v in PREDEFINED_PROMPTS.items()
|
897 |
+
],
|
898 |
label="π Predefined Scenarios",
|
899 |
+
value=list(PREDEFINED_PROMPTS.keys())[0],
|
900 |
)
|
901 |
+
|
902 |
# Prompt inputs
|
903 |
heatmap_prompt1_input = gr.Textbox(
|
904 |
label="Prompt 1",
|
905 |
placeholder="Enter first prompt...",
|
906 |
lines=2,
|
907 |
+
value=PREDEFINED_PROMPTS[
|
908 |
+
list(PREDEFINED_PROMPTS.keys())[0]
|
909 |
+
]["prompt1"],
|
910 |
)
|
911 |
heatmap_prompt2_input = gr.Textbox(
|
912 |
+
label="Prompt 2",
|
913 |
placeholder="Enter second prompt...",
|
914 |
lines=2,
|
915 |
+
value=PREDEFINED_PROMPTS[
|
916 |
+
list(PREDEFINED_PROMPTS.keys())[0]
|
917 |
+
]["prompt2"],
|
918 |
)
|
919 |
+
|
920 |
# Component type configuration
|
921 |
heatmap_component_dropdown = gr.Dropdown(
|
922 |
choices=[
|
923 |
("Attention Output", "attention_output"),
|
924 |
+
("MLP Output", "mlp_output"),
|
925 |
("Gate Projection", "gate_proj"),
|
926 |
("Up Projection", "up_proj"),
|
927 |
("Down Projection", "down_proj"),
|
928 |
+
("Input Normalization", "input_norm"),
|
929 |
],
|
930 |
label="Component Type",
|
931 |
value="attention_output",
|
932 |
+
info="Type of neural network component to analyze",
|
933 |
)
|
934 |
|
935 |
+
# Layer number configuration
|
936 |
heatmap_layer_number = gr.Number(
|
937 |
+
label="Layer Number",
|
938 |
value=7,
|
939 |
minimum=0,
|
940 |
step=1,
|
941 |
+
info="Layer index - varies by model (e.g., 0-15 for small models)",
|
942 |
)
|
943 |
+
|
944 |
# Generate button
|
945 |
+
heatmap_btn = gr.Button(
|
946 |
+
"π₯ Generate Heatmap Visualization",
|
947 |
+
variant="primary",
|
948 |
+
size="lg",
|
949 |
+
)
|
950 |
+
|
951 |
# Status output
|
952 |
heatmap_status = gr.Textbox(
|
953 |
+
label="Status",
|
954 |
value="Configure parameters and click 'Generate Heatmap Visualization'",
|
955 |
interactive=False,
|
956 |
lines=8,
|
957 |
+
max_lines=10,
|
958 |
)
|
959 |
+
|
960 |
# Right column: Results
|
961 |
with gr.Column(scale=1):
|
962 |
# Image display
|
|
|
966 |
show_label=True,
|
967 |
show_download_button=True,
|
968 |
interactive=False,
|
969 |
+
height=400,
|
970 |
)
|
971 |
+
|
972 |
# Download button (additional)
|
973 |
download_heatmap = gr.File(
|
974 |
+
label="π₯ Download Visualization", visible=False
|
|
|
975 |
)
|
976 |
# Update prompts when scenario changes for Heatmap
|
977 |
heatmap_scenario_dropdown.change(
|
978 |
load_predefined_prompts,
|
979 |
inputs=[heatmap_scenario_dropdown],
|
980 |
+
outputs=[heatmap_prompt1_input, heatmap_prompt2_input],
|
981 |
)
|
982 |
|
983 |
# Connect the real Heatmap function
|
984 |
heatmap_btn.click(
|
985 |
generate_heatmap_visualization,
|
986 |
inputs=[
|
987 |
+
model_dropdown, # Reutilizamos el selector de modelo global
|
988 |
+
custom_model_input, # Reutilizamos el campo de modelo custom global
|
989 |
heatmap_scenario_dropdown,
|
990 |
+
heatmap_prompt1_input,
|
991 |
heatmap_prompt2_input,
|
992 |
heatmap_component_dropdown,
|
993 |
+
heatmap_layer_number,
|
994 |
],
|
995 |
outputs=[heatmap_image, heatmap_status, download_heatmap],
|
996 |
+
show_progress=True,
|
997 |
)
|
998 |
# Footer
|
999 |
+
gr.Markdown(
|
1000 |
+
"""
|
1001 |
---
|
1002 |
**π How to use:**
|
1003 |
1. Check that the backend is running
|
|
|
1006 |
4. Generate visualizations to analyze potential biases
|
1007 |
|
1008 |
**π Resources:** [OptiPFair Documentation](https://github.com/peremartra/optipfair) |
|
1009 |
+
"""
|
1010 |
+
)
|
|
|
1011 |
|
1012 |
+
return interface
|