Hao Ren

MLE @ TikTok

[Sebastian Raschka @rasbt]

I (finally) put together a new LLM Architecture Gallery that collects the architecture figures all in one place!
https://t.co/NO7z6XSRHS

[Thariq @trq212]

https://t.co/nKTDfC7zMm

[Andrej Karpathy @karpathy]

New art project.
Train and inference GPT in 243 lines of pure, dependency-free Python. This is the *full* algorithmic content of what is needed. Everything else is just for efficiency. I cannot simplify this any further.
https://t.co/HmiRrQugnP

[Grant Sanderson @3blue1brown]

On a recent trip to the UK, @standupmaths shared this delightful little probability fact with me and asked if I might animate an explanation for a video of his. It's a fun one!

See his full video with shenanigans relating this to dice and more. https://t.co/DpxkVyYVFY

[François Fleuret @francoisfleuret]

ELBO people, is that reasonable?

[Reynold Xin @rxin]

Free Dolly! Introducing the first *commercially viable*, open source, instruction-following LLM. We are open-sourcing the entirety of Dolly 2.0, including the training code, the dataset, and the model weights, all suitable for commercial use.

https://t.co/EHu4TbFl6M

[Michael Black @Michael_J_Black]

PhD students, don't worry. Technologies, trends, and even whole fields come and go. A PhD makes you an expert in a field but, more importantly, teaches you how to become an expert. Once you know that you can learn anything, you can adapt to major disruptions in your field.

[James Zou @james_y_zou]

Given a trained #ML model, can we estimate how well it works on new data w/o seeing any labels?

Can we still do it even if both covariates and labels shift in data?

Our #NeurIPS2022 paper shows how to est model perf on new data in this hard case! https://t.co/VCmlNvWLjc 1/2

[NBA @NBA]

Goodnight from Boston.

[Francis Bach @BachFrancis]

If Anderson acceleration didn't speed up your algorithm, you should try Richardson extrapolation, as described in this new blog post.
https://t.co/jXwrRIp4QU

[Ting Chen @tingchenai]

Introducing SimCLR: a Simple framework for Contrastive Learning of Representations. SimCLR advances previous SOTA in self-supervised and semi-supervised learning on ImageNet by 7-10% (see next).

https://t.co/X5CXud0VwL

Joint work with @skornblith @mo_norouzi @geoffreyhinton.

[Li Haoyi @li_haoyi]

Published a new blog post:

Beyond Liskov: Type Safe Equality in #Scala

https://t.co/8RacjHecgM

[IntelliJ Scala @IntelliJScala]

Method chain hints in #IntelliJ #Scala plugin 2019.3:

[Fermat's Library @fermatslibrary]

Start with 82 and go backwards to 1

8281807978777675747372717069686 7666564636261605958575655545352 5150494847464544434241403938373 6353433323130292827262524232221 2019181716151413121110987654321

...this number is prime!

[Li Haoyi @li_haoyi]

Our #Scala by the Bay talk is out! Speedy Scala Builds with Bazel at Databricks

https://t.co/ZkYf1bMhWN

[Quoc Le @quocleix]

Want to improve accuracy and robustness of your model? Use unlabeled data!

Our new work uses self-training on unlabeled data to achieve 87.4% top-1 on ImageNet, 1% better than SOTA. Huge gains are seen on harder benchmarks (ImageNet-A, C and P).

Link: https://t.co/ZYDaef6sdp

[Lilian Weng @lilianweng]

😊Self-supervised learning opens up a huge opportunity for better utilizing unlabelled data while learning in a supervised learning manner. My latest post covers many interesting ideas of self-supervised learning tasks on images, videos & control problems: https://t.co/SdUrQQDzzy

[Uber Engineering @UberEng]

While many neural networks perform quite well at their tasks, they are fundamentally complex systems, and their training and operation is still poorly understood. https://t.co/sXIQBU6kP2

[Microsoft Research @MSFTResearch]

Microsoft Researchers with collaborators at @CarnegieMellon and @Stanford created PipeDream, a new way to parallelize deep neural network training. See how PipeDream gets 5.3x faster training time by combining intra- and inter-batch parallelism: https://t.co/8ieraUBKVr #SOSP19