Top all-time · #8251
myrtle.ai
Company blog · myrtle.ai/learnaiprogramming
myrtle.ai is the company blog at myrtle.ai, writing about ai and programming. Since 2018, 10 different Hacker News readers have linked it 11 times, most often “How to train your Resnet (79 seconds to 94%)”. No feed was found, so read it on the site itself.
Too few sampled comments (6) to say how often HN recommends it.?
Why HN links it
HN readers link Myrtle's "How to train your ResNet" series, which covers training ResNets on CIFAR-10, including learning rate, batch size, momentum, batch normalisation and tricks such as input patch whitening.
Summarised by Claude from a sample of 6 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Deploy machine learning models that run in microseconds with low latency AI inference from myrtle.ai.
Most-linked posts?
- /learn/how-to-train-your-resnet-8-bag-of-tricks/2 people · 2 mentions · first linked 2022-12-23
- How to train your Resnet (79 seconds to 94%)2 people · 2 mentions · first linked 2019-06-24
- Training ResNet9 in 26s2 people · 2 mentions · first linked 2019-08-23
What Hacker News says?
Companies like https://myrtle.ai/ and https://www.graphcore.ai/ are popping up.
Comment linking /
on “AMD is having its day”- nl · 2019-06-24
Submitted How to train your Resnet (79 seconds to 94%) as a story · 2 points
HN discussion OK, I exaggerated a little because I was recalling from memory: the old fast.ai approach actually takes <18 minutes ( https://www.fast.ai/2018/08/10/fastai-diu-imagenet/ ). My bad. (I'm sure it's improved since then but I don't know how much.) I was also thinking of https://myrtle.ai/how-to-train-your-resnet-8-bag-of-tricks/ which does CIFAR-10 in 26s but I'm not sure offhand what CIFAR-10's SOTAs look like so not s…
Comment linking /how-to-train-your-resnet-8-bag-of-tricks/
on “OpenGPT-2: We Replicated GPT-2 Because You Can Too”Submitted Vision Transformers 1: Low Earth Orbit Satellites as a story · 1 points
HN discussionYou should check out our blog post related to finding the optimal learning rate, batch size, and momentum setting: https://www.myrtle.ai/2018/09/24/how-to-train-your-resnet-5/
Comment linking /2018/09/24/how-to-train-your-resnet-5/
on “A Bayesian Perspective on Generalization and Stochastic Gradient Descent”Submitted Training ResNet9 in 26s as a story · 1 points
HN discussionMyrtle has a similar series that includes discussion about batch normalisation: https://myrtle.ai/learn/how-to-train-your-resnet/ https://twitter.com/dcpage3/status/1141700299071066112 Disclaimer: I work at Myrtle!
Comment linking /learn/how-to-train-your-resnet/
on “L2 Regularization and Batch Norm”Submitted Learn machine learning models. Practical tips and tricks as a story · 1 points
HN discussionSomeone else linked this in a different tree but I do truly love this blogpost series -- you can see the section "input patch whitening" for more details. https://myrtle.ai/learn/how-to-train-your-resnet-8-bag-of-tr... Alternatively, you can read through the direct code here if you're willing to wade through some of the mathy code on it: https://github.com/tysam-code/hlb-CIFAR10/blob/main/main.py#...
Comment linking /learn/how-to-train-your-resnet-8-bag-of-tricks/
on “Do Simpler Machine Learning Models Exist and How Can We Find Them?”Submitted How to Train Your RESNET as a story · 1 points
HN discussionNice experiment! Not to minimize your work, in fact some ideas might be complementary to yours, a couple years ago there were already some approaches reaching 30s on a single V100: https://myrtle.ai/learn/how-to-train-your-resnet-8-bag-of-tr...
Comment linking /learn/how-to-train-your-resnet-8-bag-of-tricks/
on “Do Simpler Machine Learning Models Exist and How Can We Find Them?”
