File size: 16,169 Bytes
fed9343 74acfe3 fed9343 56c210b c38002e fed9343 7077f95 fed9343 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 | import gradio as gr
import pandas as pd
import json
import os
from .dataset_logic import DatasetAgent, save_to_json, load_from_json
# Global instance for UI
agent = DatasetAgent()
def generate_blueprint_handler(topic, depth, breadth, model_name, api_key):
if not topic:
return None, None
df = agent.generate_blueprint(topic, int(depth), int(breadth), model_name=model_name, api_key=api_key or None)
filepath = save_to_json(df, "blueprint")
return df, filepath
def generate_qa_handler(blueprint_df, model_name, api_key):
if blueprint_df is None or blueprint_df.empty:
return None, None
# Agent iterates over the tags in the blueprint to generate MCQs
final_df = agent.generate_qa_batch(blueprint_df, model_name=model_name, api_key=api_key or None)
filepath = save_to_json(final_df, "full_dataset")
return final_df, filepath
def load_json_handler(file):
if file is None: return None
return load_from_json(file)
def save_json_handler(df, prefix):
if df is None: return None
return save_to_json(df, prefix)
def create_ui():
"""
Constructs the Performance-Aware Model Compression for Circuit Analysis GUI.
"""
theme = gr.themes.Soft(
primary_hue="blue",
secondary_hue="gray",
font=[gr.themes.GoogleFont("Outfit"), "sans-serif"]
)
with gr.Blocks(theme=theme, title="Model Compression Engine") as app:
gr.Markdown("""
# 🧪 Performance-Aware Model Compression for Circuit Analysis
*A workflow for building hierarchical prerequisite graphs for model compression evaluation.*
""")
with gr.Accordion("🔑 API Configuration", open=False):
api_key_input = gr.Textbox(
label="OpenAI API Key (Optional, defaults to .env)",
placeholder="sk-...",
type="password"
)
with gr.Tabs():
# PHASE 1: BLUEPRINTTab
with gr.TabItem("Phase 1: Dataset Blueprint"):
with gr.Row():
with gr.Column(scale=1):
gr.Markdown("### 1. Configuration")
model_dropdown = gr.Dropdown(
choices=["gpt-4o", "gpt-4o-mini", "gpt-4-turbo", "gpt-3.5-turbo"],
value="gpt-4o",
label="OpenAI Model for Generation"
)
topic = gr.Textbox(
label="Main Subject / Topic",
placeholder="e.g. Modern Web Infrastructure and Scalability",
lines=2
)
with gr.Row():
depth = gr.Slider(1, 5, 2, step=1, label="Graph Depth")
breadth = gr.Slider(1, 4, 2, step=1, label="Parent Fan-out")
gen_bp_btn = gr.Button("🚀 Generate Blueprint", variant="primary")
gr.Markdown("---")
gr.Markdown("### 2. File Import/Export")
upload_bp = gr.File(
label="Upload Existing Blueprint (JSON)",
file_types=[".json"]
)
gr.Examples(
examples=[["assets/blueprint_7o054luj.json"]],
inputs=upload_bp,
label="Blueprint Example"
)
download_bp_link = gr.File(label="Download Generated Blueprint", interactive=False)
with gr.Column(scale=3):
gr.Markdown("### 3. Blueprint Editor (Hierarchical Tags)")
blueprint_table = gr.Dataframe(
interactive=True,
wrap=True,
label="Edit Tags and Structure before Phase 2"
)
gen_bp_btn.click(
generate_blueprint_handler,
inputs=[topic, depth, breadth, model_dropdown, api_key_input],
outputs=[blueprint_table, download_bp_link]
)
upload_bp.change(load_json_handler, inputs=[upload_bp], outputs=[blueprint_table])
# PHASE 2: Q&A GENERATION
with gr.TabItem("Phase 2: Question & Answer Generation"):
gr.Markdown("### Submit the Blueprint for Full Q&A Generation")
gr.Info("The agent will iterate over every entry's tags and generate specific MCQs using function calling.")
with gr.Row():
run_qa_btn = gr.Button("⚡ Generate Questions & Answers", variant="primary", scale=1)
save_final_btn = gr.Button("💾 Save Final Dataset Snapshot", variant="secondary", scale=1)
with gr.Row():
with gr.Column(scale=1):
upload_final = gr.File(
label="Upload Final Dataset (JSON)",
file_types=[".json"]
)
gr.Examples(
examples=[["assets/full_dataset_6hxx81y3.json"]],
inputs=upload_final,
label="Full Dataset Example"
)
download_final_link = gr.File(label="Download Full Results", interactive=False)
with gr.Column(scale=3):
final_table = gr.Dataframe(
interactive=True,
wrap=True,
label="Final Dataset Preview (MCQ & Ground Truth)"
)
run_qa_btn.click(
generate_qa_handler,
inputs=[blueprint_table, model_dropdown, api_key_input],
outputs=[final_table, download_final_link]
)
save_final_btn.click(
save_json_handler,
inputs=[final_table, gr.State("final_dataset")],
outputs=[download_final_link]
)
upload_final.change(load_json_handler, inputs=[upload_final], outputs=[final_table])
# PHASE 3: STRATEGIC EVALUATION
with gr.TabItem("3. Strategic Evaluation"):
tag_sets_state = gr.State({}) # Stores computed sets
full_results_state = gr.State(None) # Stores raw DF for re-calc
gr.Markdown("### 🪜 Hierarchical Model Cascade & DFS Eval")
gr.Info("""
This phase runs a 'laddering' evaluation:
- Starts with the smallest model.
- Tracks **path-cumulative failures**.
- **Upgrades** to a larger model if failures exceed the threshold.
- Perpetuates the better model for all child descendants.
""")
with gr.Row():
with gr.Column(scale=1):
with gr.Group():
gr.Markdown("#### Evaluation Config")
provider_selection = gr.Radio(
choices=["OpenAI", "Ollama"],
value="Ollama",
label="Model Provider"
)
model_list = gr.Textbox(
label="Sorted Model List (Smallest to Largest)",
value="gemma3:270m, gemma3:1b, gemma3:4b",
placeholder="e.g. gpt-4o-mini, gpt-4o"
)
model_list_examples = gr.Examples(
examples=[["gpt-4o-mini, gpt-4o"],
["gemma3:270m, gemma3:1b, gemma3:4b"]],
inputs=model_list,
label="Model List Example"
)
fail_threshold = gr.Slider(0, 5, 1, step=1, label="Path Failure Threshold (Upgrade Trigger)")
run_eval_btn = gr.Button("🔎 Start Strategic Evaluation", variant="primary")
gr.Markdown("---")
gr.Markdown("#### Output Summary")
eval_summary_out = gr.Markdown("No evaluation run yet.")
eval_download_link = gr.File(label="Download Final Report", interactive=False)
with gr.Column(scale=3):
gr.Markdown("---")
upload_eval_report = gr.File(label="Upload Existing Eval Report (JSON)", file_types=[".json"])
with gr.Row():
eval_report_example_file_path = "assets/full_eval_report_2u970ry9.json"
gr.Examples(
examples=[eval_report_example_file_path],
inputs=upload_eval_report,
label="Eval Report Example"
)
example_btn_uploader = gr.Button("Load Example")
example_btn_uploader.click(
fn = lambda: eval_report_example_file_path,
inputs=[],
outputs=upload_eval_report
)
with gr.Tabs():
with gr.TabItem("📈 Graph Visualization"):
eval_viz_out = gr.Image(label="Dependency & Model Usage Map", type="numpy")
with gr.TabItem("🏷️ Tag Set Analysis"):
with gr.Row():
set_selector = gr.Dropdown(
choices=[],
multiselect=True,
max_choices=3,
label="Select Sets to Compare (Min 2, Max 3)"
)
update_venn_btn = gr.Button("🔄 Update Venn Diagram")
with gr.Row():
monotonic_checkbox = gr.Checkbox(
label="Apply Monotonic Capability Assumption (Small → Large inheritance)",
value=False
)
with gr.Row():
venn_img_out = gr.Image(label="Venn Diagram", type="numpy")
venn_report_out = gr.Markdown("Select sets and click update.")
with gr.TabItem("📋 Full Results Table"):
eval_table_out = gr.Dataframe(interactive=False, wrap=False)
def run_strategic_eval_handler(df, provider, models_str, threshold, api_key):
from .eval_logic import EvaluationEngine
if df is None or df.empty or not models_str:
return "Please provide dataset.", None, None, None, {}, gr.update()
# 1. Parse Models
models = [m.strip() for m in models_str.split(',')]
# 2. Run Engine
engine = EvaluationEngine(models, int(threshold), provider=provider, api_key=api_key or None)
results_df = engine.run_cascade_eval(df)
# 3. Finalize results
summary = engine.generate_summary(results_df)
viz_img = engine.generate_visualization(results_df)
report_path = save_to_json(results_df, "full_eval_report")
tag_sets = engine.get_tag_sets(results_df)
available_sets = list(tag_sets.keys())
import numpy as np
from PIL import Image
import io
img = Image.open(io.BytesIO(viz_img))
return summary, report_path, np.array(img), results_df, tag_sets, gr.update(choices=available_sets, value=available_sets[:3] if len(available_sets) >=3 else available_sets)
def load_eval_report_handler(file):
from .eval_logic import EvaluationEngine # load_from_json is already imported in this file and is not available in .eval_logic
if file is None: return None, None, None, None, {}, gr.update(choices=[], value=[])
results_df = load_from_json(file)
if results_df.empty:
return "Error loading file.", None, None, None, {}, gr.update(choices=[], value=[])
# Instantiate engine with defaults just for viz/summary mapping
engine = EvaluationEngine([], 0)
summary = engine.generate_summary(results_df)
viz_img = engine.generate_visualization(results_df)
tag_sets = engine.get_tag_sets(results_df)
available_sets = list(tag_sets.keys())
import numpy as np
from PIL import Image
import io
img = Image.open(io.BytesIO(viz_img))
return summary, file.name, np.array(img), results_df, tag_sets, gr.update(choices=available_sets, value=available_sets[:3] if len(available_sets) >=3 else available_sets)
def update_venn_handler(results_df, selected, apply_monotonic):
from .eval_logic import EvaluationEngine
if results_df is None or len(selected) < 2:
return None, "Select at least 2 sets."
engine = EvaluationEngine([], 0)
tag_sets = engine.get_tag_sets(results_df, apply_monotonic=apply_monotonic)
viz_bytes, report = engine.plot_tag_venn(tag_sets, selected)
import numpy as np
from PIL import Image
import io
img = Image.open(io.BytesIO(viz_bytes))
return np.array(img), report
run_eval_btn.click(
run_strategic_eval_handler,
inputs=[final_table, provider_selection, model_list, fail_threshold, api_key_input],
outputs=[eval_summary_out, eval_download_link, eval_viz_out, eval_table_out, tag_sets_state, set_selector]
).then(
lambda df: df, inputs=[eval_table_out], outputs=[full_results_state]
)
upload_eval_report.change(
load_eval_report_handler,
inputs=[upload_eval_report],
outputs=[eval_summary_out, eval_download_link, eval_viz_out, eval_table_out, tag_sets_state, set_selector]
).then(
lambda df: df, inputs=[eval_table_out], outputs=[full_results_state]
)
update_venn_btn.click(
update_venn_handler,
inputs=[full_results_state, set_selector, monotonic_checkbox],
outputs=[venn_img_out, venn_report_out]
)
return app
|