David

🐭🦀

[Everlier @Everlier]

@nvidia New RTX "Pay to Play" :)

[蔡子博士Chris @caiziboshi]

川普1314
普京520
太会选日子了！

[Manya Koetse @manyapan]

On social media, Trump's China visit has already started 😂

[琉璃 @rxliuli]

如果是在中国开发 GitHub 的话（为了避免某司说我故意污名化，我重做了一张

[Saganism @Saganismm]

"If we are honest — and scientists have to be — we must admit that religion is a jumble of false assertions, with no basis in reality. The very idea of God is a product of the human imagination. It is quite understandable why primitive people, who were so much more exposed to the overpowering forces of nature than we are today, should have personified these forces in fear and trembling. But nowadays, when we understand so many natural processes, we have no need for such solutions. I can't for the life of me see how the postulate of an Almighty God helps us in any way."

— Paul Dirac, Remarks made during the Fifth Solvay International Conference

[David @davidshen84]

@ghosTM55 这个“你说得对”包含了几层含义啊！😂

[Overleaf @overleaf]

TikZ is a powerful package to create graphic elements in LaTeX! We've written a article that introduces some basic concepts (drawing lines, dots, curves, circles, etc) to get you started. https://t.co/UhEdZ43kjP

[山谷剛史 アジア中国ITライター＆異国飯 @YamayaT]

立ち乗り二輪車でキョンシーという発想・・・！

[刘江/LIU Jiang @turingbook]

赞。智谱和MiniMax上市而且获得较高市值后，我相信更多中国的大模型实验室现在更有资源和心气儿搞一些基础研究了。

[GoGavin @GoGavin22]

@CattardSlim @TLangley99 https://t.co/3K2Xw3HpcX

[Databricks AI Research @DbrxMosaicAI]

New research from Databricks AI Research: FlashOptim cuts training memory by over 50% with no measurable loss in model quality.

Training a model with AdamW typically requires 16 bytes per parameter just for weights, gradients, and optimizer state. FlashOptim brings that down to 7 bytes, or 5 with gradient release. For Llama-3.1-8B finetuning, peak GPU memory drops from 175 GiB to 113 GiB.

Two techniques drive this: improved master weight splitting using tighter ULP-normalized error correction, and companded optimizer state quantization that reduces quantization error and improves convergence.

FlashOptim works as a drop-in replacement for SGD, AdamW, and Lion, supports distributed training with DDP and FSDP2, and is open source.

Paper: https://t.co/TBaRh90g8k
Source code: https://t.co/Dh6vOsSIbh

[The Great Translation Movement 大翻译运动 @TGTM_Official]

This is a horse.

[David @davidshen84]

"We called him Tortoise because he taught us," 🐢

[David @davidshen84]

The more there is of mine, the less there is of yours.

#Duchess

[David @davidshen84]

#NieRAutomata is so good 👍 really should have a sequel

[蔡英文 Tsai Ing-wen @iingwen]

Merry Christmas!
Wishing all those celebrating in Taiwan & around the world a joyful holiday spent with loved ones.

[西乔 XiQiao @recatm]

噶了

[David @davidshen84]

Look how much I learned on Duolingo in 2025! How did you do? #Duolingo365