5 ms·

I ran this for most of today in the background and I have some thoughts: The quality is good and it's giving you all of the maps (as well as the .blends!). It

by avaer 3y ago

I ran this for most of today in the background and I have some thoughts: The quality is good and it's giving you all of the maps (as well as the .blends!). It seems great for its stated goal of generating ground truth for training. However, it's very slow/CPU bound (go get lunch) so probably doesn't make sense for applications with users behind the computer in the current state. Additionally, the .blend files are so unoptimized that you can't even edit them on a laptop with texturing on. The larger generations will OOM a single run on a reasonably beefy server. To be fair, these warnings are in the documentation. With some optimization (of the output) you could probably do some cool things with the resulting assets, but I would agree with the authors the best use case is where you need a full image set (diffuse, depth, segmentation) for training, where you can run this for a week on a cluster. To hype this up as No Man's Sky is a stretch (NMS is a marvel in its own right, but has a completely different set of tradeoffs). EDIT: Although there are configuration files you can use to create your own "biomes", there is no easy way to control this with an LLM. Maybe you might be able to hack GPT-4 functions to get the right format for it to be accepted, but I wouldn't expect great results from that technique.

justinclift 3y ago
> ... on a reasonably beefy server. Out of curiosity, what kind of cpu / ram are you meaning here? Asking because I have some spare hardware just sitting around, so am thinking... :)
- p1esk 3y ago
  A typical server would be 8xA100 with a dual CPU (128 cores total), 2TB of RAM. I doubt you have something like this sitting around.
  - justinclift 3y ago
    That's only a "typical server" for some companies with specialised needs. Lots of places have servers with 128-256GB of ram around though.
    - bragr 3y ago
      That doesn't track in my experience, but that depends heavily on how you define typical. Just going off total servers installed, what I see getting racked are typically 1+TB RAM, and anything less would be seen as low density (i.e. cost inefficient). We've got a whole batch in the 512GB range that are coming up on EOL. Dual socket is definitely less common, but not rare either.
      - _delirium 3y ago
        In my corner of academia, 128gb is by far the most common RAM per node on something billed as a compute cluster (on random desktops and laptops it’s of course much lower than 128gb). I have seen a few 1tb+ nodes but they are rare.
      - CSMastermind 3y ago
        I know nothing about server hardware but I'm curious how that works. I have a decent PC (AMD 3990X 64-Core Processor with 256 GB of RAM), I'd have installed better/more components but that seemed to be the best you could do on the consumer market a few years ago when I was building it. Are they using the same RAM I'm using with a different motherboard that just supports more of it? Or are they using different components entirely? Apologies for what I'm sure is a very basic question but it would be interesting to learn about.
        dur-randir 3y ago
        Typical Intel offering is 1.5-2TB per socket. Socket scales up to 8 (though the price increase 2->4 is very steep). Memory itself is registered ECC DIMMs (which is even lower cost than consumer DIMMs/unbuffered ECCs), but to get to 1.5TB density you need low-rank (LRDIMM) modules, which gives x2 capacity but at a higher price.
        elif 3y ago
        It's the same RAM chips (though error tolerance features are prioritized over pure speed). You would just need a server motherboard to support that many sockets, and a server chassis to support that motherboard, and a rack to support the cooling needs of that chassis. Here's what a lowly 256GB server looks like. For a TB just imagine even more sticks: https://i.ebayimg.com/images/g/dnIAAOSwcy1kFNqq/s-l1200.jpg https://i.ebayimg.com/images/g/dnIAAOSwcy1kFNqq/s-l1200.jpg
fnordpiglet 3y ago
Proof of concepts don’t have to be optimized. That’s an exercise for the reader. ;-)

by avaer 3y ago

justinclift 3y ago

> ... on a reasonably beefy server. Out of curiosity, what kind of cpu / ram are you meaning here? Asking because I have some spare hardware just sitting around, so am thinking... :)

p1esk 3y ago
A typical server would be 8xA100 with a dual CPU (128 cores total), 2TB of RAM. I doubt you have something like this sitting around.
- justinclift 3y ago
  That's only a "typical server" for some companies with specialised needs. Lots of places have servers with 128-256GB of ram around though.
  - bragr 3y ago
    That doesn't track in my experience, but that depends heavily on how you define typical. Just going off total servers installed, what I see getting racked are typically 1+TB RAM, and anything less would be seen as low density (i.e. cost inefficient). We've got a whole batch in the 512GB range that are coming up on EOL. Dual socket is definitely less common, but not rare either.
    - _delirium 3y ago
      In my corner of academia, 128gb is by far the most common RAM per node on something billed as a compute cluster (on random desktops and laptops it’s of course much lower than 128gb). I have seen a few 1tb+ nodes but they are rare.
    - CSMastermind 3y ago
      I know nothing about server hardware but I'm curious how that works. I have a decent PC (AMD 3990X 64-Core Processor with 256 GB of RAM), I'd have installed better/more components but that seemed to be the best you could do on the consumer market a few years ago when I was building it. Are they using the same RAM I'm using with a different motherboard that just supports more of it? Or are they using different components entirely? Apologies for what I'm sure is a very basic question but it would be interesting to learn about.
      - dur-randir 3y ago
        Typical Intel offering is 1.5-2TB per socket. Socket scales up to 8 (though the price increase 2->4 is very steep). Memory itself is registered ECC DIMMs (which is even lower cost than consumer DIMMs/unbuffered ECCs), but to get to 1.5TB density you need low-rank (LRDIMM) modules, which gives x2 capacity but at a higher price.
      - elif 3y ago
        It's the same RAM chips (though error tolerance features are prioritized over pure speed). You would just need a server motherboard to support that many sockets, and a server chassis to support that motherboard, and a rack to support the cooling needs of that chassis. Here's what a lowly 256GB server looks like. For a TB just imagine even more sticks: https://i.ebayimg.com/images/g/dnIAAOSwcy1kFNqq/s-l1200.jpg https://i.ebayimg.com/images/g/dnIAAOSwcy1kFNqq/s-l1200.jpg

fnordpiglet 3y ago

Proof of concepts don’t have to be optimized. That’s an exercise for the reader. ;-)