Semblance

@coinbureau @BTW is this an AI Video

🚨ANTHROPIC CEO: OPEN SOURCE AI IS GETTING DANGEROUS Anthropic CEO Dario Amodei told lawmakers that open-source AI is moving down a “very dangerous path.” His warns that once powerful models are released openly, companies lose the ability to monitor misuse, revoke access, or update safety guardrails.

@coinbureau maybe training on free data was thieving i think he is in desperate need of money, ,you can only terrorize up-to an extent ,and after that it feels like any normal day

@sakurayukiai has anyone implemented this

the write-up that kicked this off: arxiv.org/abs/2604.26557

Offloading your KV cache to NVMe usually just makes the OS page cache thrash your RAM. DUAL-BLADE fixes this by completely bypassing the filesystem to stream tensors straight from raw SSD blocks. I love when the solution is just ignoring the kernel ✨

@shrav_10 in his voice:-The first lets you experiment with quick results. The second comes after you've curated enough successful experiments into training data. At that point, you fine-tune the model so it learns those patterns, rather than depending on retrieval every time.

Interviewed another candidate today. I asked him what RAG is. He replied: RAG is a technique that allows an LLM to answer questions using external knowledge instead of relying only on what it learned during training. Then I asked about fine-tuning. He said: Fine-tuning is the process of training a pre-trained language model on your own dataset so that it learns a new behavior, style, or task. Then I asked a follow-up: Since RAG can inject external knowledge and even mimic a user's persona through retrieved context, why do we still need fine-tuning? When would you choose one over the other? What would your answer be?

@also_priyanka sharing is caring right!!!

@piratedmango Ya

tired? well your competition isn't

@levidiamode @blelbach @GPU_MODE back to basics 🤖

@levidiamode @blelbach @GPU_MODE so here is my question ,wont this put you on a loop for trying and fixing one solution after the next ,idk it feels like that to me ,or are you guys giving it a step by step instructions ..

171/365 of GPU Programming Finally got below 2ms on the GPU Mode QR challenge. Still quite far away from CUDA colonel @blelbach, so I'm excited to read everyones' writeups once the competition is over. I especially wonder how #1 is utilizing NVFP4 (or is the submission name a false flag?). Every lower precision attempt has not panned out for me so far unfortunately. Very curious where the top 3 will converge in a week from now... If you're still debating whether to participate, would highly recommend! You learn so much from just trying stuff and being in the discord. And @modal gives you $30 in monthly compute for free. Maybe we can get a bit more compute for the next challenge if we all ask @charles_irl nicely 😁😁

@AnshitGupta7x7 I AM BATMAN

I’m 21 - bought myself lego(hackathon money) - no rich parents - not “got lucky” - no IIT - set a goal and achieved it Discipline, Consistency, Priorities✌🏻

@AshwinN14729359 BLUE J

What language did you write your very first program in?

@Taniyatweets_ @avrldotdev he is rusting

@avrldotdev 512MB? Bro is coding on pure nostalgia

Be honest, for a dev, how much RAM is enough? 1. 4 GB 2. 8 GB 3. 16 GB 4. 24 GB 5. 32 GB 6. ⁠64 GB

@Taniyatweets_ for a mac,128gb no doubt ,in this day and age get apple care too

@Zinny_Edmund it helps u gain speed .test how it improved your system then u wont need to do ,redo it

Does a developer necessarily have to know Data structure and Algorithms?

@aadityansha_06 think about the matrices as a whole ,do u really need that much of data transfers :- p.s:-Linear algebra

Looking for sponsors for my upcoming article about "How I optimized 1024*1024 matrix from 58 sec to less than 0.2 sec on my i3 2nd gen outdated processor"

@shivanijpatel Jealous

hardware people, how do you keep your desk neat?

@Rishikesh04code medium articles are dope , they simplify a lot of things

exams first then back to building ( tbh the beauty of CSE <3 never gets old )

@karanbuidls @rohan_jvm if u still want a read , will drop u on DM's

@karanbuidls @rohan_jvm the incomplete draft? ,its like not even touching the basics ,so u can really understand what's going on and secondly it doesn't describe data's chracteristic. like:-is the matrix symmetric or not 2ndly-if its diagonal dependent or dependent on the off diagonal one 3rd r99

Thanking my 5 year younger self today :) Back in 2021, got myself an RTX 3050 & now it’s giving me a 13.5x speedup on 1000x1000 matrix multiplication Now it’s time to see what it can do for deep learning :)

@_arohan_ SVD - has its limitations u need to be actually a chef to cook the whole thing

Whats your favorite linear algebra operation? My colleague said SVD must be made by god himself because it’s a miracle in 2d.