SoftMax @DataGod_v1
In God We trust all others must bring data and memes San Francisco, CA Joined June 2022-
Tweets577
-
Followers429
-
Following889
-
Likes614
How does @deepseek_ai Sparse Attention (DSA) work? It has 2 components: the Lightning Indexer and Sparse Multi-Latent Attention (MLA). The indexer keeps a small key cache of 128 per token (vs. 512 for MLA). It scores incoming queries. The top-2048 tokens to pass to Sparse MLA.
How does @deepseek_ai Sparse Attention (DSA) work? It has 2 components: the Lightning Indexer and Sparse Multi-Latent Attention (MLA). The indexer keeps a small key cache of 128 per token (vs. 512 for MLA). It scores incoming queries. The top-2048 tokens to pass to Sparse MLA. https://t.co/QzzPRvAaNa
New in-depth blog post time: "Inside NVIDIA GPUs: Anatomy of high performance matmul kernels". If you want to deeply understand how one writes state of the art matmul kernels in CUDA read along. (Remember matmul is the single most important operation that transformers execute…
A good, light read on hardware basics such as cache, prefetch, false sharing, and branches. needoneapp.medium.com/the-hardware-k…
(1/6) Ever wondered how GPUs efficiently access data for computing? 🤔 In Triton, the magic is in tl.make_block_ptr. I wrote a blog covering: - how tensors live in memory - make_block_ptr (+ striding & offset) - and more about ML and Triton kernels It's short with visuals! 🧵
Tri Dao (creator of FlashAttention) says there are 3 kinds of inference we will need to optimize for: > traditional chatbot workloads w/ fast enough to feel responsive but not instantaneous, to maintain a natural user experience > low-latency ultra-fast inference for highly…
🚀 Introducing Qwen3-Omni — the first natively end-to-end omni-modal AI unifying text, image, audio & video in one model — no modality trade-offs! 🏆 SOTA on 22/36 audio & AV benchmarks 🌍 119L text / 19L speech in / 10L speech out ⚡ 211ms latency | 🎧 30-min audio…
Some perf related must-reads: • How to Optimize a CUDA Matmul Kernel for cuBLAS-like Performance: siboehm.com/articles/22/CU… • Outperforming cuBLAS on H100: a Worklog: cudaforfun.substack.com/p/outperformin… • Defeating Nondeterminism in LLM Inference: thinkingmachines.ai/blog/defeating… • Making Deep…
Congrats to @deepseek_ai ! DeepSeek-R1 was published in Nature yesterday as the cover article, and vLLM is proud to have supported its RL training and inference🥰
This is an insanely large world created using our 3D world generation model. It blew my mind!
This is an insanely large world created using our 3D world generation model. It blew my mind!
Introducing Bunny - world's first curiosity device for kids It’s screenfree..it’s portable.. We raised $1M from @southpkcommons to reimagine how kids thrive in the age of AI, safely. Comment 'Bunny'. Our nephew will pick 50 families that get it for free this holiday season…
good read on the **economics** of retrieval, and study of the new AWS S3 Vectors, from a vectordb VP Eng. some learnings: > Turbopuffer on s3 is $0.33/gb - but the new S3 Vectors is now $0.06/gb. "That’s more than a 10x reduction compared to traditional vector databases." > AI…
good read on the **economics** of retrieval, and study of the new AWS S3 Vectors, from a vectordb VP Eng. some learnings: > Turbopuffer on s3 is $0.33/gb - but the new S3 Vectors is now $0.06/gb. "That’s more than a 10x reduction compared to traditional vector databases." > AI… https://t.co/LA5P8wWT9p
Deep dive into optimizing weight transfer step by step and improving it 60x!
Deep dive into optimizing weight transfer step by step and improving it 60x!
You DO NOT want to miss this - All the tricks and optimisations used to make gpt-oss blazingly fast, all of it - in a blogpost (with benchmarks)! 🔥 We cover details ranging from MXFP4 quantisation to, pre-built kernels, Tensor/ Expert Parallelism, Continuous Batching and much…
Wow, thanks to @charles_irl , you can understand internals of vLLM with a live notebook from @modal 🥰
Wow, thanks to @charles_irl , you can understand internals of vLLM with a live notebook from @modal 🥰
quite an interesting read
KV cache compression techniques ▪️KV caching (basic) – stores previously computed Keys and Values in memory and calculates attention only for new tokens. ▪️ Quantization – represents KV cache with fewer bits. ▪️ Low-rank decomposition – compresses the KV cache into smaller…
(1/N) How close are we to enabling robots to solve the long-horizon, complex tasks that matter in everyday life? 🚨 We are thrilled to invite you to join the 1st BEHAVIOR Challenge @NeurIPS 2025, submission deadline: 11/15. 🏆 Prizes: 🥇 $1,000 🥈 $500 🥉 $300
New in-depth blog post - "Inside vLLM: Anatomy of a High-Throughput LLM Inference System". Probably the most in depth explanation of how LLM inference engines and vLLM in particular work! Took me a while to get this level of understanding of the codebase and then to write up…
GPT-OSS bug fixes + Flex Attention support is here! 1. Fixed float16 infinite losses (>65504 overflows) 2. SWA=128 Flex default uses 129 tokens (extra 1) 3. Fixed MXFP4 inference swiglu_limit=7.0 not set 4. Sink token moved to index 0 5. FA3 doesn't have attn sink dX Details:…
GPT-OSS bug fixes + Flex Attention support is here! 1. Fixed float16 infinite losses (>65504 overflows) 2. SWA=128 Flex default uses 129 tokens (extra 1) 3. Fixed MXFP4 inference swiglu_limit=7.0 not set 4. Sink token moved to index 0 5. FA3 doesn't have attn sink dX Details:… https://t.co/rUbvvjGW7W
Ciqoox @Ciqoox63052
16 Followers 2K Following
EvangelineBess @Ju7CCT75R9VbvX
7 Followers 919 Following
Ovrixea @Ovrixea062043
83 Followers 3K Following
SPAC_Tracker🇺🇸 @Florcoo883832
42 Followers 2K Following 15-30% Monthly | 2 High-Conviction Stocks.Short-Term Gains: 15-20% in Days/Weeks.DM "JOIN" for WhatsApp Alerts. Live Trade Signals • Market Analysis
Yasmin @869i3I8gHN4uxQo
12 Followers 878 Following
Big Wisky (Amen, TWMA... @JonnyCasino999
761 Followers 7K Following Geronimo was the leader of a Native American fighting force that captivated the US. His spirit lives on today in each of you. God Bless & Amen🇺🇸🏴☠️
neo @stankneo
953 Followers 5K Following Cyberpunk Metamodernism. Aspiring hyperwrangler. Searching for lcm(∞-axia). CS ∪ CogSci ∪ Complex Systems.
DJ Goosen @dj_goosen
343 Followers 178 Following Cofounder/CTO https://t.co/mtEtKWMn61 | Agentic AI, ML/DL, @Ansible, occasionally music | ex-Ticketmaster
Adaline Olson @AOlson69714
38 Followers 2K Following
Ycoujaf @Ycoujaf137
18 Followers 963 Following
Darshith V @DarshithV25205
78 Followers 191 Following Software Intern - Wabtec Corporation, https://t.co/TZjH2qWgI4 in Software Engineering - RV College, Bangalore
Lelook @Lelook2360
40 Followers 2K Following
Priorjaw @Priorjaw262942
33 Followers 1K Following
bun.bun.🐽 @ds_bun_
19K Followers 6K Following love #pugs, lead data scientist @datafying my tweet = data science, machine learning, ai, deep learning and pug as well.
Anibal Pfannerstill @AnibalPfan83392
82 Followers 3K Following
Srespear @Srespear746
41 Followers 2K Following
Resmoon @ResmoonJnJawjV
22 Followers 588 Following
Thote @ThoteTIkuhj
25 Followers 855 Following
Thuesisto @Thuesisto3Z6
34 Followers 967 Following
👑 Goddess Hunny �... @kreamiebunny89
16 Followers 125 Following I was created to be your everything 🥰 🐷 tribute: $50
Doydee @Doydee4etsEC
24 Followers 744 Following
Thetha @Thetha954748
51 Followers 2K Following
CelestialWitch324 @Srajir2271
3 Followers 174 Following
AgnesHuxley @fH1gQRN4SZS62Kf
83 Followers 2K Following
Loatou @Loatouocx7
14 Followers 281 Following
Stosecr @Stosecrfm_S
33 Followers 4K Following
Thares @TharesHJC
51 Followers 4K Following
Smeighth @SmeighthnSBJzN
96 Followers 4K Following
Tratairsm @TratairsmR7IO9
47 Followers 4K Following
Marynel11 @Marryera61
66 Followers 805 Following
AIformedicine @ai4medicine4
561 Followers 7K Following
Smeatob @smeatob58589
66 Followers 7K Following
Will M @ipadicWillma
1 Followers 165 Following
Shirley @Dnoaner8xGmtV
28 Followers 3K Following
Trairt @TrairtbkIxju
37 Followers 2K Following
Gamethoughs @gamethough9070
71 Followers 7K Following A strong woman is one who is determined to do what others are determined not to do.
Shirley @VoretheuhIOf
42 Followers 3K Following
Tytatteigh @TytatteighdBP
27 Followers 2K Following
Crear @CrearZsinFi
35 Followers 4K Following
Shirley @SoasmarexbMrcj
32 Followers 3K Following
ちゃーめいこ @chameiko196307
47 Followers 3K Following 15. The sun washes your face, the morning breeze brushes your teeth, smile, and cheer yourself up.🍓🍓
Bojan Tunguz @tunguz
253K Followers 8K Following ML ex Nvidia. Creator of @trainxgb. Data Scientist. Physicist. Catholic. Husband. Father. Stanford Alum. Memelord. e/xgb. AMDG.
Kirk Borne @KirkDBorne
471K Followers 6K Following Advisor to startups. Freelancer. Founder of @LeadershipData. Global Speaker. Top influencer #BigData #DataScience #AI #IoT #ML #B2B. PhD Astrophysics @Caltech
Dan | Machine Learnin... @DanKornas
85K Followers 501 Following End-to-End ML Engineer. Building the best AI learning resource at https://t.co/lC2UKMtRjj. Youtube: https://t.co/pjpX8NvUn5
Charly Wargnier @DataChaz
139K Followers 45K Following Ex @Streamlit @Snowflake Maestro 🪄 • X about AI agents, LLMs, web apps, Python & SEO • My ❤️ is open source • DM for collabs 📩
abhishek @abhi1thakur
94K Followers 1K Following AI and ML, ex-Hugging Face, World's First 4x GM @kaggle, YouTube 100k+: https://t.co/BHnem8fTu5
Scott Gray @scottgray76
9K Followers 794 Following GPU Geek at @OpenAI. I have a long standing interest in neuroscience and its application to machine learning. He/Him.
Wall Street Apes @WallStreetApes
1.2M Followers 31K Following We Are The Resistance. Unfiltered Breaking News | Followed By @elonmusk 𝕏 @joerogan 🎙️ @DonaldjTrumpJr 🇺🇸 @dbongino ⚖️ @RealAlexJones 🪬 @JamesOKeefeIII 🗞️
Rona likes compilers @ronawang
31K Followers 614 Following compiler engineer (please hire me) // @mit math & cs
Stéphane Liem Nguyen @stephliemnguyen
31 Followers 102 Following PhD student in Machine Learning at @UNIGE_en
Daniel Han @danielhanchen
28K Followers 2K Following Building @UnslothAI. Faster RL / training. LLMs bug hunter. OSS package https://t.co/aRyAAgKOR7. YC S24. Prev ML at NVIDIA. Hyperlearn used by NASA.
Ruiqi Gao @RuiqiGao
9K Followers 783 Following Research scientist @GoogleDeepmind | Generative models. Veo3, Veo2, CAT3D, Imagen Video, etc. | Mom of Mochi.
Lilian Weng @lilianweng
167K Followers 167 Following Co-founder of Thinking Machines Lab @thinkymachines; Ex-VP, AI Safety & robotics, applied research @OpenAI; Author of Lil'Log
Taco Cohen @TacoCohen
27K Followers 3K Following Post-trainologer at FAIR. Into codegen, RL, equivariance, generative models. Spent time at Qualcomm, Scyfer (acquired), UvA, Deepmind, OpenAI.
Mr John C @Mister_John_C
25 Followers 566 Following
Merty @mertologico11
707 Followers 8K Following growth specialist @appgco | MVP in 10 Days • Fix My Design
Krishna Kumar @krishnakumar_nn
8 Followers 129 Following interested in AI, Education, Economics, Politics, Evolution, ...
Ashutosh Kumar @ashu_1069
699 Followers 945 Following fixing software of self-driving cars while stalking physics like Kohli on 99
DJ Goosen @dj_goosen
343 Followers 178 Following Cofounder/CTO https://t.co/mtEtKWMn61 | Agentic AI, ML/DL, @Ansible, occasionally music | ex-Ticketmaster
Alfredo Canziani @alfcnz
119K Followers 296 Following Musician, math lover, cook, dancer, 🏳️🌈, and an ass prof of Computer Science at New York University
Jim Jimson 🦙 @jimjimson_
1K Followers 4K Following Teacher of Geoffrey Hinton. Founder of AI. Retired geneticist. Software developer. Mechatronics engineer. Man of action.
neo @stankneo
953 Followers 5K Following Cyberpunk Metamodernism. Aspiring hyperwrangler. Searching for lcm(∞-axia). CS ∪ CogSci ∪ Complex Systems.
François Fleuret @francoisfleuret
46K Followers 487 Following Research Scientist @meta (FAIR), Prof. @Unige_en, co-founder @nc_shape. I like reality.
Evan @StockMKTNewz
642K Followers 396 Following Free Stock Market News that is FAST, ACCURATE, CONSISTENT, and RELIABLE | Not Just Stock News | My Daily Stock Market Recap is the link in my bio ⬇️
Mario Souto @mariohsouto
253 Followers 356 Following Building AI for energy @ stealth startup / ex-AWS Energy
Jimmy Apples 🍎/acc @apples_jimmy
59K Followers 2K Following Wagmi. 2025. As featured in Bloomberg. As quoted by Nobel Prize winner Demis Hassabis. As mentioned on the Lex Fridman Podcast💺
jason liu @jxnlco
43K Followers 2K Following independent ai consultant, a16z scout, creator of instructor prev. @stitchfix @meta
Bindu Reddy @bindureddy
165K Followers 326 Following CEO of @abacusai, the world’s first AI super assistant and general-purpose agent, DeepAgent, for enterprises and professionals. ex-GM, AWS and Google
Taseen @tntaseen
246 Followers 256 Following
DIRTY DAN @captdirtydan
410 Followers 1K Following Full-time shitposter, part-time shipbuilding enjoyer.
Manzoor Strange @_realmanzoor
168 Followers 628 Following Full-stack dev dropping fire with JS, C++, Python, React & Next.js. Into AI, gaming & apps for good. Wanna build something sick? DM me!
Katsu @Katsuu_9
135 Followers 1K Following
dunce fundz (L4 RTRD ... @DunceFundz
623 Followers 988 Following LoveSavestheDay $L4 ! 💥 Blood in blood out 🩸 GET 5 sol (.25 minimum) for FREE at the link below
nikhil tayal @Alloutnikhil
3K Followers 5K Following I love to build products and services that people want to use. Trying my best to be max useful to the humanity.
John Allan @JohnFAllan
584 Followers 2K Following e/acc + exec search partner + ai startup co-founder
robert mine @imrobertmine
1K Followers 871 Following
Alex Kehr @alexkehr
29K Followers 5K Following ceo, @superlocalmaps (acq by foursquare) • i like maps, design, and making apps (@machineofideas)
Inverse Gary Marcus �... @InverseMarcus
597 Followers 3K Following Professional Goal-Post mover. Parody account.
flowstate @k_flowstate
4K Followers 701 Following still looking for a C healer ❤️🩹 | ALX grad 🎓 | AI Aficionado 🤖
Citizen Lane @laneshetron
333 Followers 839 Following swe @aws / ☀️ founder / math & econ @columbia ♔ | prev: @wbd, @StockX
Chris Chambless will ... @Lumenbeing
196 Followers 281 Following coincidence theorist debunker, fact checker checker
GosuCoder @GosuCoder
4K Followers 75 Following Programmer that Loves to Share Things on YouTube especially about AI and new technology! Building https://t.co/mEanyLBqsATrends for United States
You might like
