-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathdata.ts
More file actions
150 lines (141 loc) · 4.39 KB
/
data.ts
File metadata and controls
150 lines (141 loc) · 4.39 KB
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
// Site taxonomy + playground tool registry.
// Article counts are computed from the posts collection at build time — see
// countPostsByTopic() below.
export type Topic = {
id: string;
name: string;
desc: string;
};
export const TOPICS: Topic[] = [
{
id: 'inference',
name: 'Inference & Serving',
desc: 'vLLM, TGI, paged attention, continuous batching, speculative decoding.',
},
{
id: 'training',
name: 'Training Systems',
desc: 'Trainers, optimizers, recipes, debugging large runs.',
},
{
id: 'architecture',
name: 'Architecture',
desc: "Transformers, MoE, SSMs, hybrids, and what's next.",
},
{
id: 'distributed',
name: 'Distributed Training',
desc: 'FSDP, tensor parallel, pipeline parallel, sequence parallel.',
},
{
id: 'quantization',
name: 'Quantization',
desc: 'PTQ, QAT, FP4, FP8, mixed precision, calibration.',
},
{
id: 'rag',
name: 'Retrieval & RAG',
desc: 'Embeddings, indexes, re-rankers, and pipeline systems.',
},
{
id: 'multimodal',
name: 'Models',
desc: 'LLMs, VLMs, multimodal systems, capabilities, and model behavior.',
},
{
id: 'agents',
name: 'Agents',
desc: 'Planning, tool use, multi-agent systems, memory, and orchestration.',
},
{
id: 'evals',
name: 'Evaluation',
desc: 'Benchmarks, harnesses, contamination, signal vs noise.',
},
{
id: 'mlops',
name: 'MLOps & Deployment',
desc: 'Pipelines, monitoring, observability, regressions.',
},
];
export function countPostsByTopic<T extends { data: { topicId: string } }>(
posts: T[],
): Record<string, number> {
const counts: Record<string, number> = {};
for (const t of TOPICS) counts[t.id] = 0;
for (const p of posts) {
if (counts[p.data.topicId] !== undefined) counts[p.data.topicId]++;
}
return counts;
}
export type ExternalToolCategory =
| 'Tokenization'
| 'Memory & VRAM'
| 'Architecture'
| 'Training & Scaling';
export type ExternalTool = {
name: string;
source: string;
desc: string;
href: string;
category: ExternalToolCategory;
};
export const EXTERNAL_TOOLS: ExternalTool[] = [
{
name: 'The Tokenizer Playground',
source: 'Xenova · Hugging Face',
desc: 'Compare how GPT-4, LLaMA, Mistral, Qwen, Gemma, and others tokenize the same text — side by side, in the browser.',
href: 'https://huggingface.co/spaces/Xenova/the-tokenizer-playground',
category: 'Tokenization',
},
{
name: 'Tiktokenizer',
source: 'dqbd',
desc: 'OpenAI-focused tokenizer playground. Visualize cl100k, o200k, and legacy encodings with per-token highlights.',
href: 'https://tiktokenizer.vercel.app/',
category: 'Tokenization',
},
{
name: 'LLM Model VRAM Calculator',
source: 'NyxKrage · Hugging Face',
desc: 'Widely-referenced inference VRAM estimator for popular open-source models with quantization and context-length sliders.',
href: 'https://huggingface.co/spaces/NyxKrage/LLM-Model-VRAM-Calculator',
category: 'Memory & VRAM',
},
{
name: 'APXML VRAM Calculator',
source: 'APXML',
desc: 'Inference-focused VRAM calculator covering Nvidia GPUs and Apple Silicon. Good for picking hardware for a target model.',
href: 'https://apxml.com/tools/vram-calculator',
category: 'Memory & VRAM',
},
{
name: 'LLM Visualization',
source: 'Brendan Bycroft',
desc: 'A 3D, animated walk through the entire forward pass of GPT-2 nano, layer by layer. The clearest mental model of how a transformer works.',
href: 'https://bbycroft.net/llm',
category: 'Architecture',
},
{
name: 'Chinchilla Scaling Calculator',
source: 'Nathan Godey',
desc: 'Plug in a compute budget, get the compute-optimal model and data size per Hoffmann et al. 2022. Charts the iso-loss surface too.',
href: 'https://nathangodey.github.io/posts/scaling/',
category: 'Training & Scaling',
},
];
export function formatDate(iso: string): string {
const d = new Date(iso);
return d.toLocaleDateString('en-US', { month: 'short', day: '2-digit', year: 'numeric' });
}
export function sortPostsByDate<T extends { id: string; data: { date: Date } }>(
a: T,
b: T,
): number {
const d = +b.data.date - +a.data.date;
return d !== 0 ? d : a.id.localeCompare(b.id);
}
export function formatMonth(iso: string): string {
const d = new Date(iso);
return d.toLocaleDateString('en-US', { month: 'long', year: 'numeric' });
}