Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
File size: 15,213 Bytes
c6c8587 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 |
/**
* Decision Path 3D Visualization - Fixed Version
*
* Shows the exact path through the neural network when generating tokens,
* highlighting critical layers, attention patterns, and decision factors.
* This is the core "Glass Box" visualization for the PhD thesis.
*
* @component
*/
"use client";
import { useRef, useState, useEffect, useMemo } from "react";
import { Canvas, useFrame } from "@react-three/fiber";
import { OrbitControls } from "@react-three/drei";
import * as THREE from "three";
import {
GitBranch,
Activity,
Sparkles,
Zap,
Brain
} from "lucide-react";
interface LayerActivation {
layer_index: number;
attention_weights: number[][];
hidden_state_norm: number;
ffn_activation: number;
top_attention_heads: number[];
confidence: number;
}
interface DecisionPath {
token: string;
token_id: number;
probability: number;
layer_activations: LayerActivation[];
attention_flow: Array<{
from_layer: number;
to_layer: number | string;
strength: number;
top_heads: number[];
}>;
alternatives: Array<{
token: string;
token_id: number;
probability: number;
}>;
decision_factors: {
attention_focus: number;
semantic_alignment: number;
syntactic_correctness: number;
context_relevance: number;
confidence: number;
};
critical_layers: number[];
confidence_score: number;
timestamp: number;
}
// Enhanced Layer Component with proper FFN visualization
interface LayerProps {
position: [number, number, number];
layerIndex: number;
isCritical: boolean;
isActive: boolean;
activation?: LayerActivation;
}
function Layer({ position, layerIndex, isCritical, isActive, activation }: LayerProps) {
const meshRef = useRef<THREE.Mesh>(null);
const ffnRef = useRef<THREE.Mesh>(null);
useFrame((state) => {
if (meshRef.current && isCritical) {
const scale = 1 + Math.sin(state.clock.elapsedTime * 3) * 0.1;
meshRef.current.scale.set(scale, scale, scale);
}
if (ffnRef.current && activation) {
// Pulse FFN based on activation strength
const ffnScale = 1 + (activation.ffn_activation * 0.2);
ffnRef.current.scale.set(1, ffnScale, 1);
}
});
const baseColor = isCritical ? "#ff6b6b" : isActive ? "#4ecdc4" : "#2d3748";
const ffnColor = isCritical ? "#e91e63" : isActive ? "#9c27b0" : "#6b46c1";
return (
<group position={position}>
{/* Main attention layer */}
<mesh ref={meshRef}>
<boxGeometry args={[4, 0.3, 2]} />
<meshStandardMaterial
color={baseColor}
emissive={isCritical ? baseColor : "#000000"}
emissiveIntensity={isCritical ? 0.3 : 0}
metalness={0.6}
roughness={0.3}
/>
</mesh>
{/* FFN Component - positioned behind */}
<group position={[0, 0, -1.5]}>
<mesh ref={ffnRef}>
<boxGeometry args={[3, 0.2, 0.8]} />
<meshStandardMaterial
color={ffnColor}
emissive={ffnColor}
emissiveIntensity={isActive ? 0.2 : 0.1}
metalness={0.7}
roughness={0.3}
/>
</mesh>
</group>
{/* Attention heads visualization - small cubes */}
{isActive && (
<group position={[0, 0, 1.2]}>
{Array.from({ length: 16 }).map((_, i) => (
<mesh key={i} position={[(i % 4 - 1.5) * 0.3, 0, Math.floor(i / 4) * 0.2 - 0.3]}>
<boxGeometry args={[0.15, 0.1, 0.15]} />
<meshStandardMaterial
color={activation?.top_attention_heads?.includes(i) ? "#ffd93d" : "#4a5568"}
emissive={activation?.top_attention_heads?.includes(i) ? "#ffd93d" : "#000000"}
emissiveIntensity={activation?.top_attention_heads?.includes(i) ? 0.5 : 0}
/>
</mesh>
))}
</group>
)}
</group>
);
}
// Simple scene
function DecisionPathScene({ decisionPath }: { decisionPath: DecisionPath | null }) {
const numLayers = 20;
const layerSpacing = 3.5;
return (
<>
<ambientLight intensity={0.5} />
<pointLight position={[10, 10, 10]} intensity={1} />
<directionalLight position={[0, 10, 5]} intensity={0.5} />
{/* Input Layer */}
<mesh position={[0, -5, 0]}>
<boxGeometry args={[5, 0.2, 2]} />
<meshStandardMaterial color="#10b981" />
</mesh>
{/* Transformer Layers */}
{Array.from({ length: numLayers }).map((_, i) => {
const isCritical = decisionPath?.critical_layers?.includes(i) || false;
const activation = decisionPath?.layer_activations?.[i];
return (
<Layer
key={i}
position={[0, i * layerSpacing, 0]}
layerIndex={i}
isCritical={isCritical}
isActive={!!activation}
activation={activation}
/>
);
})}
{/* Output Layer */}
<mesh position={[0, numLayers * layerSpacing + 5, 0]}>
<boxGeometry args={[5, 0.2, 2]} />
<meshStandardMaterial color="#f59e0b" />
</mesh>
{/* Connection lines - simplified for now */}
{decisionPath && decisionPath.critical_layers && decisionPath.critical_layers.map((layerIdx, idx) => {
const startY = layerIdx * layerSpacing;
const endY = idx < decisionPath.critical_layers.length - 1
? decisionPath.critical_layers[idx + 1] * layerSpacing
: numLayers * layerSpacing + 5;
const points = [];
points.push(new THREE.Vector3(0, startY, 0));
points.push(new THREE.Vector3(0, endY, 0));
const geometry = new THREE.BufferGeometry().setFromPoints(points);
return (
<primitive
key={`line-${idx}`}
object={new THREE.Line(
geometry,
new THREE.LineBasicMaterial({ color: 0xff6b6b, linewidth: 3 })
)}
/>
);
})}
<gridHelper args={[100, 100, 0x444444, 0x222222]} />
</>
);
}
export default function DecisionPath3DFixed() {
const [decisionPath, setDecisionPath] = useState<DecisionPath | null>(null);
const [isConnected, setIsConnected] = useState(false);
const [isAnalyzing, setIsAnalyzing] = useState(false);
const [mounted, setMounted] = useState(false);
const [modelLoading, setModelLoading] = useState(true);
const [loadingProgress, setLoadingProgress] = useState(0);
const [loadingMessage, setLoadingMessage] = useState("Initializing...");
const wsRef = useRef<WebSocket | null>(null);
const [prompt, setPrompt] = useState("def quicksort(arr):");
useEffect(() => {
setMounted(true);
}, []);
useEffect(() => {
if (!mounted) return;
const connectToService = () => {
try {
const ws = new WebSocket('ws://localhost:8769');
ws.onopen = () => {
console.log('[DecisionPath3D] Connected to service');
setIsConnected(true);
wsRef.current = ws;
// Don't immediately set as ready - wait for model_ready or loading_progress messages
};
ws.onmessage = (event) => {
console.log('[DecisionPath3D] Raw message received:', event.data);
const data = JSON.parse(event.data);
console.log('[DecisionPath3D] Parsed message type:', data.type);
console.log('[DecisionPath3D] Message data:', data);
if (data.type === 'decision_path') {
console.log('[DecisionPath3D] Setting decision path with critical layers:', data.data?.critical_layers);
setDecisionPath(data.data);
} else if (data.type === 'analysis_complete') {
console.log('[DecisionPath3D] Analysis complete');
setIsAnalyzing(false);
} else if (data.type === 'loading_progress') {
setLoadingProgress(data.progress);
setLoadingMessage(data.message);
if (data.progress === 100) {
setModelLoading(false);
}
} else if (data.type === 'model_ready') {
setModelLoading(false);
setLoadingProgress(100);
setLoadingMessage("Model ready!");
} else if (data.type === 'loading_error') {
setModelLoading(false);
setLoadingMessage(`Error: ${data.message}`);
}
};
ws.onerror = (error) => {
console.log('[DecisionPath3D] Service not available');
setIsConnected(false);
};
ws.onclose = () => {
console.log('[DecisionPath3D] Disconnected');
setIsConnected(false);
wsRef.current = null;
};
} catch (error) {
console.log('[DecisionPath3D] Connection failed');
setIsConnected(false);
}
};
connectToService();
return () => {
if (wsRef.current) {
wsRef.current.close();
}
};
}, [mounted]);
const startAnalysis = () => {
console.log('[DecisionPath3D] Start analysis clicked');
console.log('[DecisionPath3D] WebSocket state:', wsRef.current?.readyState);
console.log('[DecisionPath3D] Is connected:', isConnected);
if (wsRef.current && wsRef.current.readyState === WebSocket.OPEN) {
console.log('[DecisionPath3D] Sending analyze request with prompt:', prompt);
setIsAnalyzing(true);
wsRef.current.send(JSON.stringify({
type: 'analyze',
prompt: prompt
}));
} else {
console.log('[DecisionPath3D] WebSocket not ready, state:', wsRef.current?.readyState);
}
};
if (!mounted) {
return (
<div className="bg-gray-900 rounded-xl p-6 h-[900px]">
<div className="flex items-center justify-center h-full">
<div className="text-gray-400">Loading 3D visualization...</div>
</div>
</div>
);
}
return (
<div className="bg-gray-900 rounded-xl p-6 h-[900px]">
{/* Header */}
<div className="flex items-center justify-between mb-4">
<div>
<h2 className="text-2xl font-bold flex items-center gap-2">
<GitBranch className="w-6 h-6 text-yellow-400" />
Decision Path Visualization
</h2>
<p className="text-gray-400 mt-1">
See exactly how the model makes its decisions - the Glass Box view
</p>
</div>
<div className="flex items-center gap-4">
<div className={`flex items-center gap-2 px-3 py-1 rounded-full ${
isConnected ? 'bg-green-900/30 text-green-400' : 'bg-yellow-900/30 text-yellow-400'
}`}>
<Activity className={`w-4 h-4 ${isConnected ? 'animate-pulse' : ''}`} />
{isConnected ? 'Connected' : 'Disconnected'}
</div>
</div>
</div>
{/* Controls */}
<div className="bg-gray-800 rounded-lg p-4 mb-4">
<div className="flex items-center gap-4">
<input
type="text"
value={prompt}
onChange={(e) => setPrompt(e.target.value)}
className="flex-1 px-3 py-2 bg-gray-900 text-white rounded-lg border border-gray-700 focus:border-blue-500 focus:outline-none font-mono text-sm"
placeholder="Enter code to analyze..."
/>
<button
onClick={startAnalysis}
disabled={!isConnected || isAnalyzing}
className="px-6 py-2 bg-yellow-600 text-white rounded-lg hover:bg-yellow-700 transition-colors disabled:opacity-50 flex items-center gap-2"
>
{isAnalyzing ? (
<>
<Activity className="w-4 h-4 animate-spin" />
Analyzing...
</>
) : (
<>
<Sparkles className="w-4 h-4" />
Analyze Decision Path
</>
)}
</button>
</div>
</div>
{/* 3D Canvas */}
<div className="h-[700px] bg-black rounded-lg relative">
{modelLoading ? (
<div className="flex flex-col items-center justify-center h-full">
<div className="text-white mb-4">
<Brain className="w-16 h-16 animate-pulse" />
</div>
<div className="text-xl text-white mb-2">Loading Model</div>
<div className="text-sm text-gray-400 mb-4">{loadingMessage}</div>
<div className="w-64 h-2 bg-gray-700 rounded-full overflow-hidden">
<div
className="h-full bg-gradient-to-r from-blue-500 to-purple-500 transition-all duration-500"
style={{ width: `${loadingProgress}%` }}
/>
</div>
<div className="text-xs text-gray-500 mt-2">{loadingProgress}%</div>
<div className="text-xs text-gray-500 mt-4">356M parameters • 20 layers • 16 attention heads</div>
</div>
) : (
<Canvas camera={{ position: [-40, 50, 40], fov: 50 }}>
<DecisionPathScene decisionPath={decisionPath} />
<OrbitControls
enablePan={true}
enableZoom={true}
enableRotate={true}
target={[0, 35, 0]}
/>
</Canvas>
)}
{/* Legend */}
<div className="absolute top-4 right-4 bg-gray-800/90 backdrop-blur rounded-lg p-3 text-xs">
<div className="font-semibold text-white mb-2">Decision Path</div>
<div className="space-y-1">
<div className="flex items-center gap-2">
<div className="w-3 h-3 bg-red-500 rounded"></div>
<span className="text-gray-300">Critical Layers</span>
</div>
<div className="flex items-center gap-2">
<div className="w-3 h-3 bg-teal-500 rounded"></div>
<span className="text-gray-300">Active Layers</span>
</div>
<div className="flex items-center gap-2">
<div className="w-3 h-3 bg-yellow-500 rounded"></div>
<span className="text-gray-300">Top Attention Heads</span>
</div>
<div className="flex items-center gap-2">
<Zap className="w-3 h-3 text-yellow-400" />
<span className="text-gray-300">Decision Flow</span>
</div>
</div>
</div>
{/* Info Panel */}
{decisionPath && (
<div className="absolute bottom-4 left-4 bg-gray-800/90 backdrop-blur rounded-lg p-3 text-xs max-w-xs">
<div className="font-semibold text-white mb-2">Current Decision</div>
<div className="space-y-1 text-gray-300">
<div>Token: <span className="text-yellow-400">{decisionPath.token}</span></div>
<div>Confidence: <span className="text-green-400">{(decisionPath.confidence_score * 100).toFixed(0)}%</span></div>
<div>Critical Layers: <span className="text-red-400">{decisionPath.critical_layers.join(", ")}</span></div>
</div>
</div>
)}
</div>
</div>
);
} |