Chris Krempel
@nudelbrot
Followers
145
Following
2K
Media
209
Statuses
2K
I'm hacking transformers on arch btw.
Berlin
Joined October 2009
POV: you’re 20km away from the capital of the 3rd largest economy in the world. When will anything ever be fixed?
1
0
1
What's the current best bang4buck to buy a desktop GPU machine that at least matches A6000 in memory bandwidth (768 GB/sec). Mac studios are only at 500 GB/sec, NV Spark even lower.
2
0
2
obligatory license note:.Imagerey licensed as dl-de/by-2-0, creator: Geoportal Berlin / Digitale farbige Orthophotos 2024 (DOP20RGBI).
0
0
0
Working on a traffic flow model and realizing how beautiful aerial imagery of our cities is.
3
0
2
Creeepy 🤯.
タコにヒントを得た対数螺旋型マニピュレータ.多種多様な物体を扱うことができる. #manipulator #gripper #RoboticHand #biorobotics #biomimicry #バイオミミクリー #生物模倣 #octopus #SpiRobs
1
0
1
Having worked through countless solutions to speed up data processing (w/o renting more HW) I'm happy to see a more structured approach coming from META. "Introducing SPDL: Faster AI model training with thread-based data loading"
1
0
0
Fast - high learning rate - pre training runs over many epochs (45) at low batch sizes (16k tok) can give you a pretty good estimate of how your actual slow, high batch size - 0.5 million tok -run will behave. Pre train. GPT2/3 medium here. The preview runs are only 15 min each.
0
0
1
Dissatisfied with google search I'm using language models to go through ~50k articles to find market research companies. I find hundreds, without any ads, without any tracking, without reading through websites with *very* custom extraction posibilities. (google for compar.)
2
0
0
I remember this: “The internet first became available on cell phones in 1996, but before affordable data plans, accidentally clicking the browser icon on your flip phone […] could cost you as much as a cheeseburger“. (Link to the article in replies).
1
0
0
Anthropic is scaling Sparse Autoencoders to their Sonnet model. IMO model interpretability is one of the most exciting research directions RN. (Discl: the many unknowns are acknowledged by the field, they are most certain that they don’t understand much).
0
0
0
Using closed source AI (@OpenAI , @AnthropicAI etc) vs Open Modes is like buying a CD vs playing in a band.
0
0
1
Listen, you do not need LangSmith or a ML observability platform for tracing agents. All you need is OpenTelemetry (e.g. jaeger-all-in-one) and a few lines of python
0
0
4
