Explore tweets tagged as #simplechains
@JuliaHub_Inc
JuliaHub
3 years
5x Faster than PyTorch - Scientific Machine Learning in #Julialang: Chris Elrod, Niklas Korsbo and Chris Rackauckas have published a new blog post about Scientific Machine Learning. They use SimpleChains.jl to achieve results 5x faster than PyTorch. https://t.co/ijtiP5LAW9 #SciML
2
31
165
@JuliaLanguage
The Julia Language
3 years
You can quickly build machine learning implementations which greatly outperform other frameworks in specialized cases using the new package SimpleChains.jl! Head over to https://t.co/MCZvpDUixP to learn more #JuliaLang
0
13
57
@marcobonici
Marco Bonici
3 years
@ChrisRackauckas @BayesWatch @pumas_ai @NKorsbo I mean, this is *amazing*. I just tested SimpleChains.jl with a simple MLP, against Flux.jl, and I obtained a 5x speedup. Also, it looks that SimpleChains.jl is much more lightweight (I monitored htop while doing the benchmarks).
1
0
0
@ChrisRackauckas
Dr. Chris Rackauckas
3 years
@dionhaefner @shoyer @HochreiterSepp But anyways, the main conclusion that I've come to now is left off of the video https://t.co/kSbfdPxiOr, but is captured in this (newer) slide. It's something discussed in more detail in the release of SimpleChains.jl ( https://t.co/HiHKLVPuY4).
1
0
0
@MathSorcerer
ごまふあざらし(GomahuAzarashi)
3 years
SimpleChains.jl is awesome.
0
0
2
@JuliaHub_Inc
JuliaHub
3 years
SimpleChains is an open source pure-Julia machine learning library. When scaling to tens of thousands of parameters, it performs 5x faster than Flux or Pytorch. To learn more about its capabilities, click here https://t.co/IV32UjiwCJ #julialang #juliacon #Flux #PyTorch
0
16
74
@MathSorcerer
ごまふあざらし(GomahuAzarashi)
1 year
Lux.jl 覚えておくと Flux.jl とか SimpleChains.jl のレイヤーにあとからコンバートできる(ドキュメント上ではできると言っている)のと multi GPU 使えるから覚えておくといいかもね.(´・ω・`) #Julia言語
0
3
12
@ChrisRackauckas
Dr. Chris Rackauckas
3 years
SimpleChains.jl with SciMLSentivitiy in #sciml #julialang:
1
3
23
@iennederthal
Julienne Canarias
12 years
blingiton_ #simplechains ready for the arm party. ;-) http://t.co/Rmhn8N3vea
0
0
0
@ChrisRackauckas
Dr. Chris Rackauckas
3 years
@jordi_bolibar A mixture of ForwardDiff and Zygote. The models of this domain are not in the realm of where we'd make effective use of Enzyme since clinical Pk/Pd models are generally 10 ODEs or less. It makes heavy use of SimpleChains.jl for this reason
0
0
1
@ChrisRackauckas
Dr. Chris Rackauckas
2 years
@arbitrandomuser @cameron_pfiffer LV is fine, it'll get community updates. The core of what makes SimpleChains.jl isn't really LV but instead its caching system for non-allocating reverse mode. That's all independent of LV.
1
0
3
@arbitrandomuser
John George Francis
2 years
@cameron_pfiffer SimpleChains.jl is super chill too ; although .. with LV deprecating i dont know where its headed. @ChrisRackauckas ?
1
0
2
@sugabogam
Sugabogam
9 months
0
0
0
@NexusCollectiv
StudioNexus
2 years
@RippleXrpie Im sure banks will jump on simplechains right before the quantum age. Makes sense indeed. #NXS, light-years ahead
0
0
0