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"use client"
import React, { useState, useRef, useEffect } from 'react';
import { motion, AnimatePresence } from 'framer-motion';
import './styles.css'; // Import a CSS file for custom styles
const components = {
input: {
color: '#E6F3FF',
name: 'Input Embedding',
mathNotation: 'E(X) ∈ ℝ^(n×d_model)',
description: 'Transforms input tokens into continuous vector representations. Each token is mapped to a d_model-dimensional vector.',
formula: 'E(x_i) = W_e * x_i, where W_e ∈ ℝ^(d_model × vocab_size)',
keyPoints: [
'Converts discrete tokens to dense vectors',
'Allows the model to process input in a continuous space',
'Learned during training to capture semantic relationships'
]
},
positional: {
color: '#FFE6E6',
name: 'Positional Encoding',
mathNotation: 'PE(pos,2i) = sin(pos/10000^(2i/d_model))',
description: 'Adds information about the position of tokens in the sequence, allowing the model to utilize the order of the sequence.',
formula: 'PE(pos,2i+1) = cos(pos/10000^(2i/d_model))',
keyPoints: [
'Injects sequence order information',
'Uses sinusoidal functions for position representation',
'Allows the model to attend to relative positions'
]
},
attention: {
color: '#E6FFE6',
name: 'Multi-Head Attention',
mathNotation: 'Attention(Q,K,V) = softmax(QK^T/√d_k)V',
description: 'Allows the model to jointly attend to information from different representation subspaces at different positions.',
formula: 'MultiHead(Q,K,V) = Concat(head_1, ..., head_h)W^O',
keyPoints: [
'Captures contextual relationships between tokens',
'Uses multiple attention heads for diverse representations',
'Enables parallel computation for efficiency'
]
},
norm: {
color: '#FFE6FF',
name: 'Layer Normalization',
mathNotation: 'LN(x) = α * (x - μ) / (σ + ε) + β',
description: 'Normalizes the inputs to have zero mean and unit variance, stabilizing the learning process.',
formula: 'μ = 1/n Σx_i, σ^2 = 1/n Σ(x_i - μ)^2',
keyPoints: [
'Stabilizes the learning process',
'Allows for training of very deep networks',
'Applied after each sub-layer in the encoder and decoder'
]
},
ffn: {
color: '#FFFDE6',
name: 'Feed-Forward Network',
mathNotation: 'FFN(x) = max(0, xW_1 + b_1)W_2 + b_2',
description: 'Applies non-linear transformations to each position separately and identically.',
formula: 'ReLU(x) = max(0, x)',
keyPoints: [
'Introduces non-linearity to the model',
'Processes each position independently',
'Typically uses ReLU activation function'
]
},
output: {
color: '#E6FFF9',
name: 'Output Linear & Softmax',
mathNotation: 'P(y|x) = softmax(xW_o + b_o)',
description: 'Transforms the final hidden states into output probabilities over the vocabulary.',
formula: 'softmax(x_i) = exp(x_i) / Σexp(x_j)',
keyPoints: [
'Projects hidden states onto output vocabulary',
'Applies softmax to obtain probability distribution',
'Used for final prediction or generation'
]
},
};
const hierarchicalItems = [
{ name: 'Input Embedding', component: 'input' },
{ name: 'Positional Encoding', component: 'positional' },
{
name: 'Encoder Stack',
children: [
{ name: 'Multi-Head Attention', component: 'attention' },
{ name: 'Layer Normalization', component: 'norm' },
{ name: 'Feed-Forward Network', component: 'ffn' },
]
},
{ name: 'Output Linear & Softmax', component: 'output' },
];
const TransformerSVG = ({ components, selectedComponent, setSelectedComponent, linkPositions }) => {
const renderComponent = (x, y, width, height, component, label, mathNotation) => (
<g key={component}>
<motion.rect
x={x}
y={y}
width={width}
height={height}
fill={components[component]?.color || "#f0f0f0"}
stroke={selectedComponent === component ? "#000" : "#666"}
strokeWidth={selectedComponent === component ? "3" : "1.5"}
rx="5"
initial={{ opacity: 0 }}
animate={{ opacity: 1, scale: selectedComponent === component ? 1.05 : 1 }}
transition={{ duration: 0.3 }}
onClick={() => setSelectedComponent(component)}
/>
<text x={x + width / 2} y={y + height / 2 - 8} textAnchor="middle" dominantBaseline="middle" fontSize="14" fontWeight="bold">
{label}
</text>
<text x={x + width / 2} y={y + height / 2 + 12} textAnchor="middle" fontSize="12" fontStyle="italic">
{mathNotation}
</text>
</g>
);
const renderArrow = (startX, startY, endX, endY, label, isMainFlow = false) => (
<g>
<motion.line
x1={startX}
y1={startY}
x2={endX}
y2={endY}
stroke={isMainFlow ? "#000" : "#666"}
strokeWidth={isMainFlow ? "2" : "1.5"}
markerEnd={isMainFlow ? "url(#arrowhead-black)" : "url(#arrowhead)"}
initial={{ pathLength: 0 }}
animate={{ pathLength: 1 }}
transition={{ duration: 1, delay: 0.5 }}
/>
{label && (
<text x={(startX + endX) / 2} y={(startY + endY) / 2 - 5} textAnchor="middle" fontSize="12" fill="#666">
{label}
</text>
)}
</g>
);
return (
<svg viewBox="0 0 800 760" className="w-full h-auto" style={{ background: '#f8f9fa' }}>
<defs>
<marker id="arrowhead" markerWidth="10" markerHeight="7" refX="0" refY="3.5" orient="auto">
<polygon points="0 0, 10 3.5, 0 7" fill="#666" />
</marker>
<marker id="arrowhead-black" markerWidth="10" markerHeight="7" refX="0" refY="3.5" orient="auto">
<polygon points="0 0, 10 3.5, 0 7" fill="#000" />
</marker>
</defs>
<text x="400" y="30" textAnchor="middle" fontSize="24" fontWeight="bold" fill="#333">Transformer Architecture</text>
{renderComponent(50, 60, 700, 50, 'input', 'Input Sequence', 'X = (x₁, ..., xₙ) ∈ ℝ^(n×d_model)')}
{renderComponent(100, 130, 600, 50, 'positional', 'Positional Encoding', 'PE(pos,2i) = sin(pos/10000^(2i/d_model))')}
<rect x="50" y="200" width="700" height="400" fill="none" stroke="#666" strokeWidth="2" strokeDasharray="5,5" rx="10" />
<text x="70" y="230" fontSize="18" fontWeight="bold" fill="#333">Encoder Stack (×N)</text>
{renderComponent(100, 250, 600, 70, 'attention', 'Multi-Head Attention', 'Attention(Q,K,V) = softmax(QK^T/√d_k)V')}
{renderComponent(100, 340, 600, 50, 'norm', 'Add & Norm', 'LayerNorm(x + MultiHeadAttention(x))')}
{renderComponent(100, 410, 600, 70, 'ffn', 'Feed Forward Network', 'FFN(x) = max(0, xW₁ + b₁)W₂ + b₂')}
{renderComponent(100, 500, 600, 50, 'norm', 'Add & Norm', 'LayerNorm(x + FFN(x))')}
{renderComponent(50, 620, 700, 50, 'output', 'Output Linear & Softmax', 'P(y|x) = softmax(xW_o + b_o)')}
{/* Main flow arrows */}
{renderArrow(400, 110, 400, 130, undefined, true)}
{renderArrow(400, 180, 400, 250, undefined, true)}
{renderArrow(400, 320, 400, 340, undefined, true)}
{renderArrow(400, 390, 400, 410, undefined, true)}
{renderArrow(400, 480, 400, 500, undefined, true)}
{renderArrow(400, 550, 400, 620, undefined, true)}
{/* Residual connections */}
<motion.path
d="M 720 285 C 750 285, 750 365, 720 365"
fill="none"
stroke="#666"
strokeWidth="1.5"
strokeDasharray="5,5"
markerEnd="url(#arrowhead)"
initial={{ pathLength: 0 }}
animate={{ pathLength: 1 }}
transition={{ duration: 1, delay: 1 }}
/>
<motion.path
d="M 720 445 C 750 445, 750 525, 720 525"
fill="none"
stroke="#666"
strokeWidth="1.5"
strokeDasharray="5,5"
markerEnd="url(#arrowhead)"
initial={{ pathLength: 0 }}
animate={{ pathLength: 1 }}
transition={{ duration: 1, delay: 1.5 }}
/>
<text x="760" y="325" textAnchor="middle" fontSize="12" fill="#666" transform="rotate(90 760 325)">Residual</text>
<text x="760" y="485" textAnchor="middle" fontSize="12" fill="#666" transform="rotate(90 760 485)">Residual</text>
{/* Render links from sidebar items to components */}
{linkPositions.map((pos, index) => (
<motion.line
key={index}
x1={pos.startX}
y1={pos.startY}
x2={pos.endX}
y2={pos.endY}
stroke="#000"
strokeWidth="2"
initial={{ opacity: 0 }}
animate={{ opacity: 1 }}
transition={{ duration: 0.5 }}
/>
))}
</svg>
);
};
const EnhancedHierarchicalList = ({ items, selectedComponent, setSelectedComponent }) => {
const handleClick = (component) => {
setSelectedComponent(component);
};
return (
<ul className="space-y-4 text-lg">
{items.map((item, index) => (
<li key={index} className="border-l-2 border-gray-300 pl-4">
{item.component ? (
<motion.button
className={`w-full text-left py-2 px-4 rounded-lg transition duration-200 ${
selectedComponent === item.component
? 'font-semibold text-blue-700 bg-blue-100'
: 'text-gray-700 hover:bg-blue-50'
}`}
onClick={() => handleClick(item.component)}
whileHover={{ x: 5 }}
whileTap={{ scale: 0.98 }}
>
{item.name}
</motion.button>
) : (
<div className="font-bold text-xl mb-2 text-gray-800">{item.name}</div>
)}
{item.children && (
<ul className="mt-2 space-y-2 ml-4">
{item.children.map((child, childIndex) => (
<li key={childIndex}>
<motion.button
className={`w-full text-left py-1 px-3 rounded-lg transition duration-200 ${
selectedComponent === child.component
? 'font-semibold text-blue-700 bg-blue-100'
: 'text-gray-600 hover:bg-blue-50'
}`}
onClick={() => handleClick(child.component)}
whileHover={{ x: 5 }}
whileTap={{ scale: 0.98 }}
>
{child.name}
</motion.button>
</li>
))}
</ul>
)}
</li>
))}
</ul>
);
};
export default function InteractiveTransformerArchitecture() {
const [selectedComponent, setSelectedComponent] = useState(null);
const detailsRef = useRef(null);
const linkPositions = [
{ startX: 400, startY: 110, endX: 400, endY: 130 }, // Input Embedding
{ startX: 400, startY: 180, endX: 400, endY: 250 }, // Positional Encoding
{ startX: 400, startY: 320, endX: 400, endY: 340 }, // Multi-Head Attention
{ startX: 400, startY: 390, endX: 400, endY: 410 }, // Layer Normalization
{ startX: 400, startY: 480, endX: 400, endY: 500 }, // Feed Forward Network
{ startX: 400, startY: 550, endX: 400, endY: 620 }, // Output Linear & Softmax
];
useEffect(() => {
if (selectedComponent && detailsRef.current) {
detailsRef.current.scrollIntoView({ behavior: 'smooth', block: 'nearest' });
}
}, [selectedComponent]);
return (
<div className="min-h-screen bg-white p-8">
<h1 className="text-4xl font-bold text-center mb-6 text-gray-800">Transformer Architecture: A Scientific Analysis</h1>
<div className="grid grid-cols-1 lg:grid-cols-3 gap-6 mb-8">
<div className="lg:col-span-2">
<div className="aspect-w-16 aspect-h-9 mb-4">
<TransformerSVG
components={components}
selectedComponent={selectedComponent}
setSelectedComponent={setSelectedComponent}
linkPositions={linkPositions}
/>
</div>
</div>
<div>
<h2 className="text-3xl font-bold mb-4 text-gray-800">Component Analysis</h2>
<EnhancedHierarchicalList
items={hierarchicalItems}
selectedComponent={selectedComponent}
setSelectedComponent={setSelectedComponent}
/>
</div>
</div>
<AnimatePresence mode="wait">
{selectedComponent && (
<motion.div
ref={detailsRef}
key={selectedComponent}
initial={{ opacity: 0, y: 20 }}
animate={{ opacity: 1, y: 0 }}
exit={{ opacity: 0, y: -20 }}
transition={{ duration: 0.2 }}
className="mt-8 p-6 border border-gray-300 rounded-lg shadow-md"
>
<h3 className="text-3xl font-bold mb-4 text-gray-800">{components[selectedComponent].name}</h3>
<div className="grid grid-cols-1 md:grid-cols-2 gap-6">
<div>
<h4 className="text-xl font-semibold mb-2 text-gray-700">Mathematical Notation:</h4>
<p className="font-mono mb-4 p-2 bg-gray-50 rounded">{components[selectedComponent].mathNotation}</p>
<h4 className="text-xl font-semibold mb-2 text-gray-700">Description:</h4>
<p className="text-gray-600 mb-4">{components[selectedComponent].description}</p>
</div>
<div>
<h4 className="text-xl font-semibold mb-2 text-gray-700">Mathematical Formulation:</h4>
<p className="font-mono mb-4 p-2 bg-gray-50 rounded">{components[selectedComponent].formula}</p>
<h4 className="text-xl font-semibold mb-2 text-gray-700">Key Points:</h4>
<ul className="list-disc pl-5 space-y-2">
{components[selectedComponent].keyPoints.map((point, index) => (
<li key={index} className="text-gray-600">{point}</li>
))}
</ul>
</div>
</div>
</motion.div>
)}
</AnimatePresence>
</div>
);
}