test_tm7873 Profile Banner
testtm Profile
testtm

@test_tm7873

Followers
1K
Following
26K
Media
2K
Statuses
13K

Love Cats, and Love Cats 🐈 πŸ‡΅πŸ‡± want to talk? come join https://t.co/b5iZQKZ2CU Support me and Cats here: https://t.co/VCHvmK4dHt

Polska
Joined October 2021
Don't wanna be here? Send us removal request.
@test_tm7873
testtm
11 months
I just created a Ko-fi page! If you’d like to support my work (or just buy me a coffee), check it out here: Will help me and cats a lot :( and bring me one step closer to my dream of creating my very own large language model (LLM). Every bit helps β™₯️.
Tweet card summary image
ko-fi.com
Support testtm
3
2
15
@test_tm7873
testtm
36 minutes
well. i did it very well. made torch ~7.2% faster than jax. (same setup, V4-32).(e. no python graph cus it didn't wnated to generate. so there is what claude coded to show the speed)
Tweet media one
1
0
2
@test_tm7873
testtm
3 hours
Well. i did it. made Torch be very very close with Jax. and even beating it by a bit on longer run. 😁.it seems that Torch XLA. loves Bfloat16. and the speed of it is EPIC!.(jax there also run on bfloat 16)
Tweet media one
@test_tm7873
testtm
2 days
I do something wrong. or there is something wrong with PyTorch XLA on TPUs. idk. they r so slow!.(512batchsize,512seqleanght and TPUV4-32)
Tweet media one
0
0
1
@test_tm7873
testtm
3 hours
Kittens r very cool! 😺.
@asklumo
Lumo
3 hours
Lumo 1.1 is an entirely different breed of private AI kitty. βœ… Smarter with less context.βœ… More aware of recent events.βœ… Higher-quality answers.βœ… More engaging. Before you ask, yes we're working on Dark Mode 🌚. See the before & after ‡️
Tweet media one
Tweet media two
Tweet media three
Tweet media four
0
0
3
@test_tm7873
testtm
4 hours
Up to 245 TB in space. speed of 14.7 GB/s. the smaller 30/61TBs r out now. the big 122/245TBs Q4 2025.
@test_tm7873
testtm
4 hours
WTF?! man that SSD is crazy!
Tweet media one
2
0
6
@test_tm7873
testtm
4 hours
source @zephyr_z9.
0
0
0
@test_tm7873
testtm
4 hours
WTF?! man that SSD is crazy!
Tweet media one
1
0
2
@test_tm7873
testtm
20 hours
What?.atleast its in 2027. the year ili change the phones again. but man it suck. .
@TechloreInc
Techlore
2 days
🚨 Google is tightening the screws on Android freedom! Starting next year, they're blocking sideloading of "unverified" apps which requires all developers to identify themselves to Google. Goal is global 2027 rollout. This is a blow to Android's open ecosystem and user choice 🧡
Tweet media one
0
0
3
@test_tm7873
testtm
21 hours
Grok Code Fast 1 released on windsruf. Free for limited time.
Tweet media one
1
0
12
@test_tm7873
testtm
21 hours
Grok Code Fast 1. released on Windsurf and its free (for limited time) !
Tweet media one
0
0
3
@test_tm7873
testtm
22 hours
Cool!.But when ratelimits increse? 😊.
@AnthropicAI
Anthropic
22 hours
We’ve developed Claude for Chrome, where Claude works directly in your browser and takes actions on your behalf. We’re releasing it at first as a research preview to 1,000 users, so we can gather real-world insights on how it’s used.
0
0
5
@test_tm7873
testtm
1 day
Look like 2.5 family is complete.
0
0
4
@test_tm7873
testtm
1 day
Gemini 2.5 Flash native image editing is crazy!.😍It look so good!
Tweet media one
0
0
7
@test_tm7873
testtm
1 day
Not home to test the banana. But cool showcase of it by my friend :).
@chetaslua
Chetaslua
1 day
Google Just Killed 100+ startup . nano banana is an industry changer. watch this 8 second video of banana greatness
1
0
6
@test_tm7873
testtm
2 days
I do something wrong. or there is something wrong with PyTorch XLA on TPUs. idk. they r so slow!.(512batchsize,512seqleanght and TPUV4-32)
Tweet media one
1
0
3
@test_tm7873
testtm
2 days
$3,499 for Jetson Thor. look very cool. but also very expensive. πŸ˜…
Tweet media one
1
0
4
@test_tm7873
testtm
2 days
whar it will need 3 separate stuff?.or they are addons. ;-;
Tweet media one
@SemiAnalysis_
SemiAnalysis
2 days
Today at Hot Chips Meta is launching it's new puck + wrist + glasses compute platform called Orion. Orion uses "WLR" or world locked rendering. There's quite a bit of compute, and the display processor, glasses processor, application processor, and compute coprocessor all use 5nm
Tweet media one
0
0
2
@test_tm7873
testtm
2 days
Google's TPU Research Cloud is ending support for v2 and v3 quota at the end of this month. So mayby new ones will be joined to program? new ones presented? interesting.
1
0
12
@test_tm7873
testtm
2 days
Google's TPU Research Cloud is ending support for v2 and v3 quota at the end of this month.
Tweet media one
1
0
8
@test_tm7873
testtm
2 days
Meta Code World Model ?.Interesting.
@btibor91
Tibor Blaho
2 days
Is "Meta Code World Model" / "CodeGen Computational World Model" new?
Tweet media one
0
0
6
@test_tm7873
testtm
2 days
Absolute crazy price in Poland for that gpu 😿 6k zΕ‚.
Tweet media one
@test_tm7873
testtm
3 days
Oh man. 😒.Ili be never able to afford one. 😭
Tweet media one
0
0
2