https://www.youtube.com/watch?v=uVcBa6NXAbk https://www.1x.tech/discover/introducing-neo-gamma
👏30❤13👍5
Signed Vlad Lialin

Channel
@dlinnlp
On this record: Growth · Engagement · What this channel posts · Reactions · Posts · Citations · Telegram's recommendations · Referenced elsewhere · Cite this entry
11,620subscribers
-70 since we began measuring on 6 August 2026
Risers and fallers across the register · movement among entries of 10,000–31,623.
| Telegram ID | -1001188069510 |
|---|---|
| Type | Channel |
| Username | @dlinnlp |
| Created | 24 September 2018 — measured — cross-checked against a third-party dataset (ext.tg_channel) |
| First recorded | 6 August 2026 |
| Last confirmed live | 6 September 2026 |
| Measurements held | 27 |
| Confirmed unchanged | 1 time, most recently 6 September 2026 |
| On Telegram | t.me/dlinnlp |
| Measured (UTC) | Subscribers | Change |
|---|---|---|
| 6 Sept 2026, 17:01 | 11,620 | -7 |
| 4 Sept 2026, 09:19 | 11,627 | -5 |
| 2 Sept 2026, 19:08 | 11,632 | -1 |
| 1 Sept 2026, 15:36 | 11,633 | -1 |
| 31 Aug 2026, 13:13 | 11,634 | -4 |
| 30 Aug 2026, 10:15 | 11,638 | -3 |
| 29 Aug 2026, 07:17 | 11,641 | -4 |
| 28 Aug 2026, 06:33 | 11,645 | -4 |
| 27 Aug 2026, 10:03 | 11,649 | -1 |
| 26 Aug 2026, 13:06 | 11,650 | -1 |
| 25 Aug 2026, 09:56 | 11,651 | +1 |
| 24 Aug 2026, 07:07 | 11,650 | -2 |
| 22 Aug 2026, 17:37 | 11,652 | -6 |
| 21 Aug 2026, 07:12 | 11,658 | +1 |
| 20 Aug 2026, 08:24 | 11,657 | -8 |
| 19 Aug 2026, 10:49 | 11,665 | -3 |
| 17 Aug 2026, 11:22 | 11,668 | -8 |
| 15 Aug 2026, 19:28 | 11,676 | -3 |
| 14 Aug 2026, 07:58 | 11,679 | -1 |
| 12 Aug 2026, 22:33 | 11,680 | first reading |
18 posts held, back to 18 July 2024 — the reader has not yet reached the start of this channel’s public history, so older posts may sit further back, unread. Read across 46 pages of Telegram’s post history, 20 posts per page.
Nothing published in the last 30 days. ERR and ER are rolling 30-day measures, so there is nothing to compute — we hold 18 posts for this entry, the most recent from 21 February 2025. An engagement rate over an empty window would be a number about nothing.
Measured directly from 3 videos with a duration reading, out of the posts we hold for this channel — not this channel’s whole posting history, only the sample this register has actually read. An exact reading to the second, taken from the post itself rather than from Telegram’s own rounded chrome, so it carries no ≈ mark.
1,446 reactions across 17 posts, in 23 distinct kinds. The most used accounts for 39.0% of them.
| Reaction | Count | Share | Share, drawn |
|---|---|---|---|
| 🔥 | 564 | 39.0% | |
| ❤ | 299 | 20.7% | |
| 👍 | 289 | 20.0% | |
| 😁 | 103 | 7.12% | |
| ❤🔥 | 74 | 5.12% | |
| 👏 | 33 | 2.28% | |
| 🤡 | 23 | 1.59% | |
| 🥰 | 20 | 1.38% | |
| 😱 | 15 | 1.04% | |
| 🥴 | 5 | 0.346% | |
| 🆒 | 3 | 0.207% | |
| 🤔 | 3 | 0.207% | |
| 🤯 | 3 | 0.207% | |
| 👎 | 2 | 0.138% | |
| 🙏 | 2 | 0.138% | |
| ✍ | 1 | 0.069% | |
| 🍾 | 1 | 0.069% | |
| 🎉 | 1 | 0.069% | |
| 👌 | 1 | 0.069% | |
| 🙈 | 1 | 0.069% | |
| 3 further kinds | 3 | 0.207% |
No sentiment is inferred, and none should be read in. This table is ordered by count and by nothing else. Emoji do not carry stable meaning across languages or communities — 🙏 is thanks in one channel and mourning in another — so we publish which ones were pressed and how often, and pass no judgement on what an audience meant by them.
Precision. Telegram publishes reaction counts per emoji and short-forms each one — 4.34K, 1.2M — so any single kind at or above 1,000 reaches us at three significant figures, and only counts below 1,000 are exact. The shares above are ratios of those figures and carry the same error. This is also why the total here can differ slightly from a reaction total printed elsewhere on the page: both are sums of the same rounded parts, taken over samples with different edges.
Coverage. Reactions were read on 17 of the 18 sampled posts in this sample. Summed by Telegram’s own count on each post — not by adding up the per-emoji breakdown above — those same posts carry 1,446 reactions in total: the kind of figure the paragraph above means by “a reaction total printed elsewhere on the page”.
Measured over the 18 most recent posts we hold, published 18 July 2024 to 21 February 2025, using the newest reading held for each. Telegram Stars are excluded: they are a payment, not a reaction, and they have their own section.
https://www.youtube.com/watch?v=uVcBa6NXAbk https://www.1x.tech/discover/introducing-neo-gamma
👏30❤13👍5
Signed Vlad Lialin
В продолжение темы, Jay Alammar, у которого были прекрасные визуальные объяснения про работу трансформера, в сто раз лучшие оригинальной статьи, выпустил только что иллюстрированный DeepSeek-R1 https://newsletter.languagemodels.co/p/the-illustrated-deepseek-r1
❤84
Signed Vlad Lialin
Всем приветики. Давно не было постов, тк становится всё сложнее вести канал. Не буду обещать что исправлюсь, но буду постить когда есть что-то о чём другие каналы не говорят достаточно. И сегодня будут не новости (о ChatGPT Operator можете прочитать где угодно), а открытая позиция на PhD студента в моей старой лабе в UMass Lowell - Text Machine Lab. Это NLPшная позиция с довольно широким спектром того чем можно зан…
🔥66👍20❤15🫡1🙈1
Signed Vlad Lialin
Programming Massively Parallel Processors https://a.co/d/6QEiuCq Наткнулся на книгу которая кажется весьма известна в мире GPU-программирования. Она довольно детально погружается в Nvidia GPU и CUDA. В четвертом издании (2022 года) ещё и добавили современные архитектуры: Ampere (A100) и Hopper (H100). Это важно тк архитектуры довольно сильно изменились с 2016 года. Очень надеюсь просмотреть хотя бы по-диагонали и н…
👍59🙏1
Signed Vlad Lialin
Но дадут ли нобелевку по литературе за Deep Learning Book
🔥115😁97🤡17🥴5❤2👏1
Signed Vlad Lialin
Почему не стоит верить nvidia-smi “GPU utilization” arthurchiao.github.io/blog/understanding-gpu-performance/ Nvidia использует очень особый способ определения утилизации GPU. 100% означают не что девайс загружен на 100%, а что хотя бы одно ядро было использовано хотя бы чуть-чуть 100% времени за последние N (мили)секунд Очень яркий пример это примитивы синхронизации: когда вы вызываете torch.barrier GPU Utilizatio…
🔥83👍22😱11❤3✍1👏1
Signed Vlad Lialin
Soumith Chintala (создатель pytorch) выдаёт базу о том как тренироваться на 10К GPU x.com/soumithchintala/status/1841498799652708712 Оч короткий TL;DR (всем рекомендую прочитать оригинал, он не длинный) 1. Maximize batch size and GPU utilization: 3D parallelism + gradient checkpointing 1. Overlap communication, e.g. while N-1th layer is computing backward, all GPUs with an Nth layer can all-reduce 1. Optimize for y…
🔥37❤20👍9👏1
Signed Vlad Lialin
https://x.com/hughbzhang/status/1838288923656941860?s=12&t=QgBLS4SmhE8cqdYBmhrqJA
Signed Vlad Lialin
O1 mini inference scaling experiments Прикольное саммари экспериментов одного чела. Коротко: если убедить модель дольше думать (что пока что непросто) pass@1 реально будет расти лог-линейно. При этом это скорее всего не majority voting или self consistency тк эти методы упираются в потолок
🔥28❤2🤔2
Signed Vlad Lialin
OpenDuck - очень классный проект по опенсорсной (хард+софт) реимплементации диснеевского робота https://github.com/apirrone/Open_Duck_Mini Очень мило. Буду следить за ними. А вот тут они уже умеют стоять: https://x.com/antoinepirrone/status/1835679313506562502
🥰17❤6👍3
Signed Vlad Lialin
Наткнулся в Твиттере на шикарную визуализацию LLM. Как выяснилось, ей уже целый год, но для новичков это все ещё полезная штука. Кроме красивой 3D-модельки, здесь еще подробный гайд по работе каждого элемента, как говорит автор, до каждого "сложить и умножить". По архитектурам там есть GPT-2, nanoGPT, GPT-2 XL, ну и GPT-3. Ссылочка на визуализацию @ai_newz
❤🔥74🔥23👍14❤3🆒3
Signed Vlad Lialin
🍓 openai.com/index/learning-to-reason-with-llms 1. GPT-o1 это затюненая с помощью RL модель на улучшение reasoning (деталей как это сделано, конечно же нет) 1. Scaling c train-time compute (как долго делать RL) и test-time compute (как долго генерировать ответ) -- на текущих графиках никакого намёка на то чтобы модель выходила на плато 🔥 1. По сравнению с 4o на codeforces o1 получает 89 перцентиль вместо 11 1. В Ph…
🔥72👍12❤6🙏1
Signed Vlad Lialin
Showing the 12 most recent of 18 posts we hold for @dlinnlp. View and reaction counts are the latest single reading for each post, not a live figure, and a recent post is still accumulating both. A view count marked ≈ was rounded by Telegram before we ever saw it — t.me prints views in full below 1,000 and to three significant figures above, so ≈1,200,000 means somewhere between 1,150,000 and 1,249,999. Unmarked counts are exact. Text is reproduced from the public post preview and truncated for length.
Republishes
Channels on the register whose posts this channel has forwarded.
Built only from forwarded posts we have actually read, on both sides. Coverage is early and deliberately incomplete: a missing link means we have not read the post that would prove it, never that the relationship does not exist. Counts are distinct forwarded posts observed, so they only ever go up as we read more.
Named by 1 registered channel — every channel on the register whose own posts have named this one, by its current username or any other username it currently holds, merged from two separately captured readings of the same fact so a namer caught by only one of them is not missed and a namer both caught is not counted twice. A username this channel has since dropped is not matched — that handle may belong to someone else now, and crediting today’s namer to yesterday’s owner would misattribute it.
Named by
Channels on the register whose posts name this channel's handle.
Names
Channels on the register whose handles appear in this channel's posts.
A mention is a weaker signal than a forward and is counted separately for that reason — naming a channel is not republishing it, and a handle in a post body is easy to place deliberately. The post counts beside each row below are distinct posts in which the handle appeared, from posts we have read on both sides — the “Named by N registered channels” figure above is a different count, of distinct NAMING CHANNELS rather than posts, and is not the sum of the rows under it.
The reverse of the list above, and a different kind of signal. This does not require this channel to have ever been asked about directly — each row below is a channel we DID ask Telegram about, whose Telegram-generated list happened to include this one. A channel can appear here with an empty list above it, because being named by someone else’s query is independent of having been queried itself.
This channel appears in 3 seed channels' Telegram-generated recommendation lists in total. Each is Telegram’s list for THAT channel, not this one — see how this is measured.
This handle named by sources this register does not control and did not measure — each shown exactly as found, attributed by name, dated to when it was read.
Hacker News
This handle was named once in a Hacker News comment or story, via the public Algolia search API. HN comment and story text has no confirmed reuse licence, so nothing quoted from either is reproduced here — only that a mention exists, when, and by whom, with a link to read it at the source.
A live page changes as we take new readings, so a citation should name the measurement it is based on, not just the URL. The line below cites the subscriber count as measured 6 September 2026 — this entry's latest reading, not the date you are reading this.
“DL in NLP” (@dlinnlp), 11,620 subscribers as measured 6 September 2026. Telegram Register, tgregister.com/channel/dlinnlp.
Full measurement history, CC BY 4.0. Every reading this register holds for this entry, not just the latest one, as a dated, downloadable record: CSV · JSON. Free to use with attribution to tgregister.com. Each file carries its own generation timestamp, which is the figure to cite for exactly when the data was retrieved.