chiphuyen / dmls-book Goto Github PK

I spotted what seems to be a minor mistake on the collection of case studies at https://github.com/chiphuyen/dmls-book/blob/main/resources.md

The author of the spotify case study is Umesh .A Bhat.

Spotify’s Discover Weekly: How machine learning finds your new music (Umesh .A Bhat, 2017)

But when opening the blog article, the author is Sophia Ciocca

Hi,

Could someone help:

For instance page 26 - Business and ML Objectives, Which post of Eugene Yan is it? Where can I find the list of all of the resources?

Cheers

The ISBN in README is incorrect, 978-1801819312 is for Machine Learning with PyTorch and Scikit-Learn: Develop machine learning and deep learning models with Python

The correct ISBN is 978-1-09-810796-3

The last sentence on page 205 "Some companies use feature stores to ensure the consistency between the batch features used during training and the streaming features used in prediction".
Is it supposed to be "... used in production" instead?

If the median is 100 ms, half of the requests take longer than 100 ms, and half of the requests take less than 100 ms.

This formulation implies a strict lower/upper bound, which the median need not be.
Just consider a sequence where every single entry is uniformly 100 ms.
The statement is easily corrected to reflect the intended meaning by inserting 2 negations.

If the median is 100 ms, half of the requests take no longer than 100 ms, and half of the requests take no less than 100 ms.

Hello,

Thank you for the very informative book and all your amazing work!

I was going through chapter 6, page 170 (of the paperback version), paragraph 2. I think that the sentence should mention data in the first sentence instead of model.
In other words, should it be "Another problem is that spreading your data on multiple machines can cause your batch size to be very big." instead of "Another problem is that spreading your model on multiple machines can cause your batch size to be very big."?

Hi Chip and team, thank you for the excellent book on ML systems design. I was going through the book and found that there is a mistake (typo) in the Graph optimization figure on page 219. The original graph, and the optimized graphs have different kernel sizes which is incorrect. I checked the cited reference link here at nvidia and indeed the original graph has a mistake on the kernel sizes. Two of the 3x3 kernels taking input from the input node must be 1x1 and two of the three 3x3 kernels in the next layer must be 5x5 and 1x1. Because this is an optimization, I thought the kernel sizes are also being optimized but after thinking about it for a while, I figured the weight dimensions will change and the trained weights will not work if that is the case.

Just wanted to make a note here in case somebody else also notices this and also if you would like to fix it in the future editions. Thank you for the amazing work once again.