initlayers

@faisalkhan4k It is

@initlayers Quantization is such a blessing.

Quantization is everywhere these days, especially with edge AI and LLM deployment becoming more common. Probably a good time to properly understand what's actually happening under the hood instead of just calling load_in_4bit=True and moving on.

@falloficarus22 They definitely do. I got my cards ready.

@initlayers but your learnings won't be wasted, maybe in some interview they would ask you questions in depth

I'm supposed to give two presentations on my 6-month internship for semester grading. And I already know how it's going to go. "What language did you use?" "What model did you train?" "What type of CNN?" "How much accuracy?" That's it. Nobody is going to ask why the architecture works, what tradeoffs exist, what failed during experimentation, or what the actual research problem even is. I can spend months understanding fundamentals, papers, optimization, systems, and implementation details, but grading here has very little to do with depth sometimes. Maybe in my next life I'll become a teacher's pet. Sounds like a much easier optimization problem.

@shookscooks Makes sense

@initlayers Dont expect apple from a mango tree

@YuvrajS9886 Okay. Noted! I'll visit these once I am done with my prep.

I highly recommend go through the quantization course hf and then reading papers >LLM.int8() >LLM QAT >LoRA >EfficientQAT >BitNet for Starters!

@closetedpe Sorry. I was busy. I just checked.

@initlayers yo dude can u check ur dms

@yugb_twt I am following this playlist youtube.com/playlist?list=…

@initlayers Link of this video....

Starting the day with FAISS. I'm slowly starting to love this library. Everything around retrieval, indexing, and similarity search just feels so much cleaner and easier with FAISS.

@Pseudo_Sid26 This just dropped. youtu.be/gh2_PhgZGsM?si…

Honestly speaking Claude is in itself a skill to learn. Using it to full capability is like gaining a full tool knowledge. Claude skills are such a boom if used in the right manner. I am not a frontend guy but some of the skills like the - impeccable skill is go good to make minimalistic frontends. This repository has 246 production-ready Claude Code skills, plugins, and agent skills for 12 AI coding tools. Over 14.5k stars !! Use this as your skill playbook, i have been doing the same, read the article then to understand the need to make your own skills as well !!

@Rishike89765470 Yes. From the Facebook

@initlayers the Facebook research Lab ?? , vornocoi are beautifully ..

@sai_45_tech My GF bought it for her research work.

@initlayers opus 4.7 , bro is rich af

This is what my girlfriend just prompted 😭 Meanwhile I'm out here treating LLMs like overworked interns and she's out here motivating them with kindness!!

@sharpeye_wnl Guess who changed it from max to high 😃

@initlayers opus 4.7 high for this, is she a multi billionaire?

@Ronak44451 IVF-PQ is the industry standard at the moment, I guess.

@initlayers Hnsw alone is already good 👍

Starting the day with Product Quantization. Planning to follow it up with a few deep drives on PQ, IVF, and HSNW after this. That's probably enough brain damage for one day.

@yugb_twt Sure. I will do

@initlayers Do post the links to the readings you do... I will also see this video...

@sir4K_zen I am not sure. Maybe another VectorDB that's better than all the existing ones.

@initlayers PQ cadence is strong, what comes after the HSNW stretch?

@Seozilla_ai Sure

@initlayers nice start with pq, ivf, and hsnw, keep it up

@datavorous_ Yes. A couple of days is good to cover them.

@initlayers PQ-Flat and PQ-IVF in a day is still okay. HNSW-PQ would be brain damage.

@yugb_twt No structure at the moment. Reading whatever I could find.

@initlayers Are you following some structure? Like I am also planning to dive deep into RAG