Transformer (deep learning architecture): Revision history

Filter revisionsshowhide

To date:

Tag filter:

Invert selection

External tools:

For any version listed below, click on its date to view it. For more help, see Help:Page history and Help:Edit summary. (cur) = difference from current version, (prev) = difference from preceding version, m = minor edit, → = section edit, ← = automatic edit summary

curprev 23:0723:07, 15 April 2025 Maxeto0910 talk contribs 108,038 bytes 0 →Speculative decoding undo Tag: Visual edit
curprev 23:0623:06, 15 April 2025 Maxeto0910 talk contribs 108,038 bytes −3 no sentences undo Tag: Visual edit

curprev 03:1103:11, 5 April 2025 24.19.113.134 talk 108,041 bytes 0 →AI boom era: move cite site one word to the left to emphasize true actual correctness of cited claim undo Tags: Mobile edit Mobile web edit

curprev 10:0010:00, 3 April 2025 Glashio talk contribs m 108,041 bytes +54 Added 2 massively widely used LLM's to the list undo

curprev 07:2607:26, 30 March 2025 Fschwarzentruber talk contribs 107,987 bytes +139 →Feedforward network: description of the terms in the equation undo Tag: Visual edit

curprev 06:2606:26, 24 March 2025 Susurrus talk contribs 107,848 bytes +4 What is a context window? undo Tag: Visual edit

curprev 02:2602:26, 14 March 2025 Cosmia Nebula talk contribs 107,844 bytes −118 →Alternative activation functions: dedup cite undo Tag: Visual edit

curprev 04:5204:52, 5 March 2025 Shen762 talk contribs m 107,962 bytes +399 →Sublayers: added a brief discussion on residual connections. undo Tag: Visual edit

curprev 02:1802:18, 24 February 2025 MathXplore talk contribs m 107,563 bytes +9 Reverted edits by 2806:263:482:2EB:88B7:87D7:C316:E924 (talk) to last version by 2A02:C7C:BDCC:3D00:B428:A3E7:5054:999 undo Tag: Rollback

curprev 16:2416:24, 19 February 2025 2a02:c7c:bdcc:3d00:b428:a3e7:5054:999 talk 107,563 bytes −39 →AI boom era undo
curprev 16:2216:22, 19 February 2025 2a02:c7c:bdcc:3d00:b428:a3e7:5054:999 talk 107,602 bytes +206 →AI boom era undo
curprev 16:1216:12, 19 February 2025 2a02:c7c:bdcc:3d00:b428:a3e7:5054:999 talk 107,396 bytes +46 →AI boom era undo

curprev 04:1504:15, 18 February 2025 Seefooddiet talk contribs 107,350 bytes −71 TMI about tertiary topics for this article's lead undo Tag: Visual edit
curprev 04:1004:10, 18 February 2025 Seefooddiet talk contribs 107,421 bytes −17 WP:SD40; this is 39. avoid repeating part of title undo Tag: Shortdesc helper

curprev 12:3612:36, 16 February 2025 Niteskate talk contribs 107,438 bytes +1 →FlashAttention: typo undo

curprev 22:3622:36, 13 February 2025 Cosmia Nebula talk contribs m 107,437 bytes −12 →Speculative decoding undo Tag: Visual edit
curprev 22:3522:35, 13 February 2025 Cosmia Nebula talk contribs 107,449 bytes +45 →Speculative decoding: correct explanation undo Tag: Visual edit

curprev 03:4403:44, 12 February 2025 73.119.178.197 talk 107,404 bytes +1 →Parallelizing attention: Parameters are countable. Fewer (vs less) is the correct word. undo Tags: Mobile edit Mobile web edit

curprev 13:2513:25, 9 February 2025 Evercat talk contribs m 107,403 bytes +2 I think "the" works slightly better here? undo

curprev 19:1519:15, 6 February 2025 Emijrp talk contribs 107,401 bytes +45 →Further reading: Category:2017 in artificial intelligence undo

curprev 03:5003:50, 3 February 2025 Cosmia Nebula talk contribs 107,356 bytes +172 →Multi-Query Attention: fig undo Tag: Visual edit
curprev 02:4502:45, 3 February 2025 Cosmia Nebula talk contribs 107,184 bytes +121 →FlashAttention: link to Softmax_function#Numerical_algorithms undo Tag: Visual edit

curprev 08:1208:12, 2 February 2025 Cosmia Nebula talk contribs 107,063 bytes +955 MLA from DeepSeek undo Tag: Visual edit
curprev 07:1207:12, 2 February 2025 Cosmia Nebula talk contribs 106,108 bytes +1,329 →Speculative decoding: multitoken prediction undo Tag: Visual edit

curprev 00:0300:03, 31 January 2025 129.2.89.24 talk 104,779 bytes 0 →Multiheaded attention: The original is just wrong, it should be Q,K,V not X X X, look at the attention is all you need paper. undo

curprev 21:0521:05, 28 January 2025 Citation bot talk contribs 104,779 bytes +1 Alter: title, template type, journal. Add: chapter-url, chapter, authors 1-1. Removed or converted URL. Removed parameters. Some additions/deletions were parameter name changes. | Use this bot. Report bugs. | Suggested by Headbomb | #UCB_toolbar undo
curprev 04:2904:29, 28 January 2025 Cosmia Nebula talk contribs 104,778 bytes +43 →Efficient implementation: anchor undo Tag: Visual edit

curprev 21:5321:53, 27 January 2025 Cosmia Nebula talk contribs 104,735 bytes +58 →Multi-Query Attention: anchor undo Tag: Visual edit

curprev 18:3518:35, 11 January 2025 Maxeto0910 talk contribs 104,855 bytes −11 no sentences undo Tag: Visual edit

curprev 20:2220:22, 10 January 2025 Toomuchcuriosity talk contribs 104,866 bytes +1,255 →Architecture: Expanding article undo

curprev 15:0915:09, 7 January 2025 Draheinsunvale talk contribs 103,611 bytes +13 →Alternative activation functions: PaLM also uses SwiGLU undo Tags: Mobile edit Mobile web edit Advanced mobile edit

curprev 09:5609:56, 6 January 2025 Madacs talk contribs m 103,598 bytes +4 Undid revision 1267720609 by Madacs (talk) undo Tag: Undo
curprev 09:5409:54, 6 January 2025 Madacs talk contribs m 103,594 bytes −4 →Parallelizing attention: removed self link undo Tags: Reverted 2017 wikitext editor
curprev 06:1606:16, 6 January 2025 OAbot talk contribs m 103,598 bytes +36 Open access bot: arxiv updated in citation with #oabot. undo
curprev 05:2005:20, 6 January 2025 CommonsDelinker talk contribs 103,562 bytes −211 Removing Ashish_Vaswani.webp; it has been deleted from Commons by Krd because: No permission since 29 December 2024. undo

curprev 22:4422:44, 2 January 2025 Carlossjuan95 talk contribs m 103,773 bytes +1,138 Added examples in the sub-quadratic transformer section. undo Tag: Visual edit

curprev 16:0316:03, 31 December 2024 136.25.107.7 talk 102,635 bytes −2 →History: As the article states, all authors were labeled as equal contributors. undo
curprev 06:3506:35, 31 December 2024 2003:d2:bf48:de00:380c:7998:2dc7:18de talk 102,637 bytes 0 →Parallelizing attention undo

15 April 2025

5 April 2025

3 April 2025

30 March 2025

24 March 2025

14 March 2025

5 March 2025

24 February 2025

23 February 2025

19 February 2025

18 February 2025

16 February 2025

13 February 2025

12 February 2025

9 February 2025

6 February 2025

3 February 2025

2 February 2025

31 January 2025

28 January 2025

27 January 2025

12 January 2025

11 January 2025

10 January 2025

7 January 2025

6 January 2025

2 January 2025

31 December 2024