Live data from Hacker News

MiniGPT-4

minigpt-4.github.io

1–10 of 337 posts

Re: MiniGPT-4

#2
I don't know what unholy abomination they did to the scrolling, but after I increased the page font size, it just sticks to the bottom and you can't scroll up.

Re: MiniGPT-4

#3
Giving this a quick spin and I like what I'm seeing. I gave it a picture of Paolo Veronese's Lament over the Dead Christ [0] and asked what was going on in the background.

> The painting depicts the scene of the crucifixion of Jesus Christ. In the foreground, there is a group of people standing around the cross, including Mary, the mother of Jesus, and the two thieves who were crucified with him. In the background, there is a hill with a group of people standing on it, watching the crucifixion. The sky is dark and stormy, with lightning illuminating the scene. The overall mood of the painting is one of sadness and tragedy.

Then I asked for a sonnet on it:

"In this painting, we see the scene of the cross

Where Jesus is crucified, with Mary at his side

And the thieves who were crucified with him

In the background, a hill with people standing

Watching the tragic event unfold

The sky is dark and stormy, with lightning flashing

Illuminating the scene of the crucifixion

The mood is one of sadness and tragedy

As we contemplate the sacrifice of Jesus

But even in death, he shows us the way

To love and forgiveness, to live another day

And so we remember the crucifixion

And the love that Jesus showed us that day."

So...not quite what I was expecting, but it's progress!

[0] https://en.wikipedia.org/wiki/Lament_over_the_Dead_Christ_(V...

Re: MiniGPT-4

#4
I was not expecting this to be good due to the hype wave, but at least looking at the examples, they seem very impressive and worthy of the (potential) hype! I can't wait to see what new waves of these models keep coming out.

I for one will probably still stay near the fundamental side of things as all of this thrash is simply far too much for me, but it is quite nice and beautiful for me to get to see all of the variety and beauty of what is being built here. Now, to figure out a good way to integrate these seamlessly into my development workflow....I go about as painfully low-tech as one could. I might as well be using an electronic typewriter with only Ctrl+F search, lolzers....

Re: MiniGPT-4

#6
On a technical level, they're doing something really simple -- take BLIP2's ViT-L+Q-former, connect it to Vicuna-13B with a linear layer, and train just the tiny layer on some datasets of image-text pairs.

But the results are pretty amazing. It completely knocks Openflamingo && even the original blip2 models out of the park. And best of all, it arrived before OpenAI's GPT-4 Image Modality did. Real win for Open Source AI.

The repo's default inference code is kind of bad -- vicuna is loaded in fp16 so it can't fit on any consumer hardware. I created a PR on the repo to load it with int8, so hopefully by tomorrow it'll be runnable by 3090/4090 users.

I also developed a toy discord bot (https://github.com/152334H/MiniGPT-4-discord-bot) to show the model to some people, but inference is very slow so I doubt I'll be hosting it publicly.

Re: MiniGPT-4

#8
post #5

Interesting, will try it out. The queue is about 100 at the moment, with 700s of waiting.

1200s now...

Also some links don't seem to open in a new tab and if you click around while you wait you might lose your progress.

I clicked on the Video button while waiting, assuming that it would open in a new tab, and lost my place in queue.

Re: MiniGPT-4

#9
post #3

Giving this a quick spin and I like what I'm seeing. I gave it a picture of Paolo Veronese's Lament over the Dead Christ [0] and asked what was going on in the background. > The painting depicts the scene of the crucifixion of Jesus Christ. In the foreground, there is a group of people standing around the cross, including Mary, the mother of Jesus, and the two thieves who were crucified with him. In the background, t…

What did you expect?

Re: MiniGPT-4

#10
post #2

I don't know what unholy abomination they did to the scrolling, but after I increased the page font size, it just sticks to the bottom and you can't scroll up.

I’m experiencing the same thing on iOS Safari, it rapidly jumps back and forth between two positions.
Post reply on HN