🇪🇺eu/acc, compressing reality (pied piper)
Joined August 2020
- Tweets 2,404
- Following 665
- Followers 241
- Likes 85,269
132 Photos and videos
Jun 9
Thought my ImageNet model was brokens until i finally increased the CFG. CFG=[1.0, 1.5, 2.0, 2.5, 3.0, 4.0]
1
24
May 24
why arent LLM people using things like REPA? i think SRA could be usefull. aligning an early layer with a late layer
53
Apr 22
what do we think about these gpt image 2 artifacts. Do you guys think the decoder is trained as a GAN/LDM/ PixelDM
3
5
457
Apr 3
So many people completly misunderstand that answer. the person is talking about local ai and since the phone doesnt need water the argumenta makes sense. just goes straight over many peoples heads
The best argument against democracy is a 5 minute conversation with the average voter
173
Being used a footstool by a homicidal maniac probably didn’t change that libtard’s suicidal empathy views at all.
Death of the West is what they all want.
It’s absolutely perfect…
A liberal man in NYC yelling about how everyone is welcome, as a Muslim man yells "Allahu Akbar!" and uses him as a literal springboard to throw a homemade bomb.
The West’s suicidal empathy on video.
2,674
12,375
93,474
26,224,501
Mar 7
When first reading @bfl_ml Self-Flow i didnt really like it and i thought it was too close to SRA but after thinking some more about it i really like it and i have some ideas to further improve it. Implementing them right now
1
58
Mar 7
I have fully switched back to 5.2 thinking. 5.4 thinking doesnt even understand the simplest problems
43
Jan 24
This is gpt 2 level. insane to think you can get more coherent responses for a few houndred $ than from a whole human
Jan 23
"Being Somali is more than bananas & rice, it's a lot, it's uh.. it's kind of like bananas & rice"
1
120
Jan 24
I just subsribed to Claude Pro because i had a bug gpt5.2 couldnt find and Opus 4.5 zero shotted it. worth the 20$ already
2
83
13 Dec 2025
if you were using bounding boxes in 2024 you were lost but if youre still using them in 2025 youre just stupid
12 Dec 2025
I slept on yesterday and I’m just not fully sold on LiDAR. After seeing the demos Rivian did (attached), I just don’t see enough of a benefit. Thoughts?
1
77
18 Nov 2025
This Paper changes so much i thought i understood about Vision transformers
2
70
15 Nov 2025
The worst thing in ML is that some mistakes dont feel or look like mistakes and seem trivial in retrospect.
1
43