Pritam

Wannabe ML psychopath

[Pritam @Pritam_ai]

I will offer one useful state of mind which, if one adopts it, makes everything much easier: "to accept reality as it is and to try not to regret the past and try to improve the situation." - Ilya

[Simon Prince @SimonPrinceAI]

I'm gradually updating and improving chapters from Understanding Deep Learning and posting them to https://t.co/8zGsFiXjXd.

The latest update concerns how to measure the performance of the trained models.

[Archie Sengupta @archiexzzz]

all the successful entrepreneurs i have met over the years have told me the same thing: life eventually becomes a pursuit of ruthless prioritization. every useless yes is a no to something useful. you have to put yourself first. the season becomes a season of silence, of focus, of choosing yourself unapologetically. you stop explaining, you stop people-pleasing, you stop shrinking to make other people comfortable. you have to protect your time, guard your energy, say less, and do more. have insane urgency in whatever you’re doing, with a lot of no’s to a lot of things.

[Pritam @Pritam_ai]

@0xSero H100 node for 10 days please btw I wanna work with you how do I 😭😭

All the best brother

[Neeraj Ong Aggarwal @a_neeraj141]

I’m looking for interns to join my team at Mistral! We work on multimodal models end to end, from pretraining to rl

https://t.co/2RDnjfTzQE

[Pritam @Pritam_ai]

@durov France is excused 😭💔

[Pritam @Pritam_ai]

@himanshustwts Link to blog?

[NIK @ns123abc]

remember if you stay alive for the next 10 years, you might stay alive forever

[Pranjal @pranjalssh]

Megakernel blog is out now

We implement a full layer (MoE + Attention) megakernel, achieving 1100 toks/s for Muse Spark, 2.1x faster than baseline.

Nvidia's Instruction cache makes it really hard to write bigger kernels. We ended up tuning both instruction and data caches at all levels for extreme latency hiding!

With this, some warps of next operation can start running before previous operation has finished - redefining what GPU utilization means.

Full blog post: https://t.co/ubZbSQpkKl

[Mithil Vakde @evilmathkid]

I got 37% on ARC-AGI-2 Kaggle without LLMs!

100M looped transformer trained from scratch

- Likely highest score without an LLM
- Peaked at 6th place
- Beats last year's winning LLM
- Looping, added in the last week: 33% → 35% → 37%

thanks to @arcprize @fchollet @mikeknoop and @GregKamradt for the competition!

Link: https://t.co/iDBUrHhmdI
(writeup for the paper track)

[Pritam @Pritam_ai]

@KyrieBlunders Pyt profiler!

[Pritam @Pritam_ai]

@Sanyam0605 @jino_rohit Which ones

[Pritam @Pritam_ai]

@iam_sanjana06 All the best

[khazzz1c @Chonghan_Liu]

they’ve ran every test I had in mind lol.

https://t.co/0WNcH0IJw0

[Pritam @Pritam_ai]

@abhibuilds Amazing profile 🤩👏🏻

[Mario Zechner @badlogicgames]

man, who should i trust? a bunch of nerds who made linux run on apple silicon, reverse engineering that pesky gpu isa among a gazillion other things, all pre-ai.

or the guy who thinks having your agent rewrite shitty rust he never looked at in x86_64 is excellence?

[Pritam @Pritam_ai]

What is taste?

[Pritam @Pritam_ai]

@0xSero Open-source for win

[Pritam @Pritam_ai]

@Raghulchan2004 Yes dm. Thanks

[Pritam @Pritam_ai]

@stanfordnlp @Diyi_Yang @michaelryan207 @jyangballin Can't wait for cs312

[Pritam @Pritam_ai]

@Raghulchan2004 We can study together if you want