20240712

I feel a bit tired, mainly because I researched how to optimize resources in machine-learning model inference very hard. Let's relax on the weekend.

• Finished reading “Just for Fun”
  • Linus says humans’ motivation factors are survival, social order, and entertainment in this order, as we no longer fight for survival, or social order at some point, everything ends up to entertainment, i.e., doing things just for fun.
  • The idea is not very new. Actually, it resembles to Maslow’s hierarchy of needs. However, his philosophy well represents the entire Linux histroy/ecosystem.
  • The latter half of the book was not quite interesting, but it was interesting to learn his humorous attitude and his personality changes on the course. The history of the early Linux was also an enjoyable material.
• Met some new people
  • They were nice. Hope this relationship continues.
• Browsed my friend's GitHub repository, submitted 6 issues, and created 1 pull request. Wow, it's nice of me.

1992A. Only Pluses

Let's consider a simpler situation in which I get to incrase one of the three numbers by 1.

a <= b <= c

I have three options

• (a + 1) * b * c = abc + bc
• a * (b + 1) * c = abc + ac
• a * b * (c + 1) = abc + ab

Because a <= b <= c, bc >= ac >= ab. Therefore, (a + 1) * b * c is the largest option, i.e., increaseing the smallest number of three maximizing their product.

Is it guaranteed that maximizing the product for each step leads to the maximum value in the end after applying the operation 5 times? I don't know…

torch.no_grad() to disable gradient calculation and save memory.

Disabling gradient calculation is useful for inference, when you are sure that you will not call Tensor.backward(). It will reduce memory consumption for computations that would otherwise have requires_grad=True.

I tried with our project, and (only) see a slight improvement on the memory usage. Or, it's totally possible that I misunderstood something and applied it in the wrong way (I don't see many web pages on the function usage).

Also, I was browing the web pages of TorchScript, but couldn't figure it out.

Cheese 10g Protein shake 10g Sashimi 0g Tofu 0g Pork 0g Bacon Egg 10g

Total carbohydrate 30g

MUST:

• Read the chapter 3 of “Code That Fits In Your Head”

TODO:

• Update my resume with the sign up feature
• Update my resume with the document structure and AWS S3
• Update my LinkedIn summary section

index 20240711 20240713