One of the most exciting innovations of LoRA , that doesn't get the "attention" it deserves, is the ability to support just-in-time parameter swapping, enabling serving hundreds (or thousands!) of fine-tuned models from a single GPU. More in the post linkedin.com/pulse/multi-lor…
3
376
Chris Johnson retweeted
Is it next year yet!? Yes, yes, from last year's perspective it is. We have published the Call for Papers for the 2nd edition of @RecSysHR! With important dates, topics, submission details, etc.! Looking forward to your submissions 🤗. recsyshr.aau.dk/cfp/ #RecSys2022
Chris Johnson retweeted
We're looking for a machine learning leader for our @NetflixResearch team that personalizes how we display content on Netflix: artwork, trailers, metadata, and more. Learn more and apply here: jobs.netflix.com/jobs/868826
Fresh deep learning for fashion project from some former @Spotify colleagues. xitter.gq/ThreadGenius/status/92…
New blog post about our API launch medium.com/@ThreadGenius/int…
Chris Johnson retweeted
Pfft… NIPS sold out years ago. Props to artisanal, small-batch conferences like AISTATS, UAI, and COLT for keeping it real. xitter.gq/miles_brundage/status/…
Next ATX Data Meetup Wed 6pm at @Amazon, "Building Google's ML engine w @tensorflow, @kubernetesio and @apachekafka" meetup.com/austindata/events…
Optimizing ML model through cross-validation, hyper-param tuning, and regularization Vs minimizing training loss at expense of overfitting
"Optimizing Vs Maximizing". Simple concept but applicable to so many scenarios. sethgodin.typepad.com/seths_…
"How to call BS on big data". Poor data-driven decision making is more hurtful than no data-driven decision making newyorker.com/tech/elements/…
3
.@google has been consistently coming up with the most click-bait-y arxiv titles "Attention is all you need", "One Model to Learn Them All" xitter.gq/karpathy/status/876628…
"One Model To Learn Them All" another step in Google's attempt to turn all of itself into one big neural network arxiv.org/abs/1706.05137
Chris Johnson retweeted
Want to work on the ML/data infra that powers #DiscoverWeekly @Spotify? Hit me up. Scala, Scio, Spark, TensorFlow, Algebird, Shapeless, ...
Chris Johnson retweeted
Check out what we're changing in Scala Collections for 2.13: scala-lang.org/blog/2017/02/…
Any good slides out there I can use to teach monoid aggregations and reductions with @ApacheSpark? cc @holdenkarau @posco @s_ryz
1
2
Chris Johnson retweeted
Thank you for everything. My last ask is the same as my first. I'm asking you to believe—not in my ability to create change, but in yours.
Chris Johnson retweeted
Hey @BarackObama, I heard you were interested in a role at Spotify. Have you seen this one? spotify.com/us/jobs/view/o0C…
Why don't more authors do this (include code)? From "A theoretically grounded app of dropout in RNNs" arxiv.org/abs/1512.05287 cc @yaringal















