The same prompt will have this random range of output depending on the random number seed.
-
- These 18 cards, all generated from the same prompt.
- Youâd think the composition would be mixed.
- Why is this?
- Initial value is just a random value
- From there, repeated noise elimination is performed to get closer to âa distribution of images that humans can find meaning in.
- We started completely differently, so the places we arrived at were completely different.
-
- In this schematic, itâs drawn in two dimensions, but in reality there are 20,000 dimensions.
- So the âsurfaceâ is very wide (Curse of the dimension).
- Almost all distributed on the surface
- We cannot observe the distribution of âthe set of things that humans can recognize as picturesâ in 20,000 dimensional space.
- In high-dimensional space, the normal distribution is almost uniformly distributed on the hypersphere
Another Perspective
-
- This is a vertical line starting from a âcommon random number seedâ
- Side by side generated by âcommon promptsâ.
- There are people in the world who are trying to prompt trial and error with AI drawing services.
- Those who donât fix the seed and use different prompts and trial and error make this observation
- Itâs almost always a random seed that determines the picture.
- If you look at this and try to find a connection to Prompt a - eto, I donât understand the translation.
- People find meaning in random events, so some people find a false âprompt connectionâ to these ârandom seed effectsâ.
- Some people have written blog posts saying, âYou should experiment with fixing the seed.â
- Those who have experimented with fixing the seed and changing the prompt make this observation
- This looks like it would be easier to make sense of than the previous example.
- Iâm throwing away E because it doesnât paint a decent picture.â
- However, when I try the same prompt with a different seed, I get this result
- I showed it to my wife, who knew nothing about it.
- My impression is that both e are quite good.
- The top is better with A and E. I like E better.
- Below, A, D, and E are good, D>A>E in that order, E is also well drawn, but he does not like the face
- In other words, even if you âfix the seed and observe with different prompts,â you will learn âhow good or bad the prompt is in a particular seed.â
- Itâs like playing a specific random map in a random map generation game.
- Thereâs no guarantee that that do-it-yourself know-how will be useful on other maps.
- Because âgoodâ in one seed doesnât match up with another seed.
- Itâs Pessimistic Misconceptions to see a bad result by accident in a certain seed and decide, âThis prompt is no good.â
- If I tried it on more other seeds, Iâd be like, âIt was just bad the first time, but it was actually surprisingly good.â
- But they underestimate it, so they donât âtry more.â
- This makes it impossible to properly estimate the probability of success.
- In the context of reinforcement learning, we are talking about how to avoid this pessimistic misunderstanding and not put more weight on exploration (Trade-offs between use and exploration).
- Itâs like playing a specific random map in a random map generation game.
- I showed it to my wife, who knew nothing about it.
- Multiple seed multiple prompts multiply and observe.
-
- A has a good chance of being so-so.
- Sometimes E can be crazy weird, but sometimes itâs very good.
- C is a weird one.
- There is a difference in probability distribution like
- This knowledge is much better than seed-fixed, trial-and-error knowledge.
- May become meaningless in future upgrades.
- How much of it depends on the language model and how much of it reflects the structure of the language itself
- We will know as various models come out in the future.
- The cat image generation in C3: Computer Created Cats uses Thompson sampling with Bernoulli distribution as the distribution (reinforcement learning).
- What is Thompson sampling?
- Sampling from a hypothetical distribution to try the largest choice.
- Appropriate attempts are made with large variances.
- C is not useful and is automatically discarded, while A and E are subject to a new trial
- Automatically update the distribution shape with the data increased by the trial
- What is Thompson sampling?
Q: Thompson sampling, I think we need feedback on the results of the evaluation, but who is doing it?
- A: Humans do it.
- About 1,500 images are generated per day, and my wife and I look at them and label them âthis one is good, this one is bad,â and we get about 100 âgood onesâ and 1,400 âbad ones.
- I was trying to improve the prompts based on the results, which was done by humans in the beginning, but it became too much of a hassle, so I w
- I said, âIf you have this much data, you can automate it.â I automated it.
- So currently, itâs like if you like something, just click the âLikeâ button and more good stuff will appear! Q: You mean try more prompts that are tied to what you think is good?
- A: Yes.
- Iâll add more later because itâs a bit of a mish-mash of explanations.
- Written Thompson Sampling Hiring Process. Q: Is there any point in pursuing too many prompts individually?
- A: I think so.
- If you want to get a good picture, you have to pull out all the stops.
- Because txt2img is, after all, a method that starts with completely random initial values.
- If you want more control, I would have to use img2img. Q: Is it hard to find a good image if Iâm doing it just right?
- A: Well, mess.
- Without a definition of what constitutes a âgood image,â the âprobability of getting a good oneâ is unknown.
- If you pull a gacha with unknown probability, you may or may not get a good one, itâs âluckâ!
- Definition of good Q: What is the range of values a seed can take?
- One integer between 0 and 4294967296 since it is just a random number seed
This page is auto-translated from [/nishio/Stable DiffusionăŽăˇăźăă¨ăăăłăăăŽé˘äż](https://scrapbox.io/nishio/Stable DiffusionăŽăˇăźăă¨ăăăłăăăŽé˘äż) using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. Iâm very happy to spread my thought to non-Japanese readers.