HNHacker News
TopNewBestAskShowJobs

heliophobicdude

689 karma · joined December 31, 2015

GH:afloresescarcega

E-mail: alex %{dot} floresescar %{@gmail}

Curious about Gaussian Splatting.

Most recently started work on splatinit.c It initializes 3D Gaussian Splats from an image and a depth map. It's fast and outputs to a typical 3D Gaussian Splat .ply file.

https://github.com/afloresescarcega/splatinit

submissionscomments
heliophobicdude··on Splatter Image: Ultra-Fast Single-View 3D Reconstruction
Hard edges are a challenge right now.

Thinking in 2D for a second, to get a nice crispy edge, you need a long and opaque splat to mark the boundary. Sometimes the long splat could wisp off leaving fuzzy artifacts.

Take this example: https://www.shadertoy.com/view/dtSfDD

Peyman Milanfar [1] suggested using bump functions instead. Bump functions would allow you to specify cut off intervals but still make the whole function smooth and continuous (good for my gradient optimization freaks)

1: https://x.com/docmilanfar/status/1719584410348204233

heliophobicdude··on SMERF: Streamable Memory Efficient Radiance Fields
Great work!!

Question for the authors, are there opportunities, where they exist, to not use optimization or tuning methods for reconstructing a model of a scene?

We are refining efficient ways of rendering a view of a scene from these models but the scenes remain static. The scenes also take a while to reconstruct too.

Can we still achieve the great look and details of RF and GS without paying for an expensive reconstruction per instance of the scene?

Are there ways of greedily reconstructing a scene with traditional CG methods into these new representations now that they are fast to render?

Please forgive any misconceptions that I may have in advanced! We really appreciate the work y'all are advancing!

heliophobicdude··on GPS-Gaussian: Generalizable Pixel-wise 3D Gaussian Splatting for Human Synthesis
The title is missing the key word "real-time" like in the paper title.

Optimization was one of the biggest bottlenecks. This will unlock fast, reconstruction on more types of hardware, more efficiently.

GS also is a great compression technique. Even so, Meta called their avatar one a Codec.

heliophobicdude··on Gaussian Head Avatar: Ultra High-Fidelity Head Avatar via Dynamic Gaussians
This is excellent! A similar paper from Meta was published two days ago on Avatars too

https://shunsukesaito.github.io/rgca/

Here's the discussion for it(empty as of right now): https://news.ycombinator.com/item?id=38554537

heliophobicdude··on GaussianDreamer: Fast Generation from Text to 3D Gaussian Splatting
Here is the discussion for the other recent paper by some of them on 4D Gaussian Splatting

https://news.ycombinator.com/item?id=37905601

heliophobicdude··on Efficient Memory Management for Large Language Model Serving with PagedAttention
Thank you so much for the response! This is really good info!
heliophobicdude··on The Chaos at Unity
When did The Verge start paywalling?

https://archive.ph/nygAO

heliophobicdude··on Optimizing LLMs from a Dataset Perspective
Perhaps not smarter but filtering seemingly computed or non-derived answers in demonstration data. Also include demonstrations that show steps leading up to a computation in a format for a runtime to parse and execute the computation. Similar to the Code Interpreter or Advanced Data Analysis in ChatGPT.

Taking it a step further, I would include in the demonstration a test harness set up with a test suite to prove the proposed implementation.

I would go through each demonstration which a fixed set of criteria measuring only passing tests but ones that also show a level of complexity and usefulness.

Why? I was looking through CodeLlamas demonstration data for fine tuning and saw answers that were not even checked for correctness or usefulness.

heliophobicdude··on Efficient Memory Management for Large Language Model Serving with PagedAttention
Ah, I see. This isn't necessarily virtualizing the static weights but the variable -sized and data dependent key value caches. These caches are built up as you go through the sequence of tokens. Makes sense.

How doesn't paging worsen speed performance though? If you are making more trips to the memory, then are you really just saving vram?

Also I see that vLLM which implements PagedAttention is also using a better scheduling? Wouldn't the speed improvements be coming from that instead? Don't put an expected short input and output in the same batch as a big input and big output?

What are the results of using the sequence-length only without virtualization?

heliophobicdude··on Video-Game Company Unity Closes Offices Following Death Threat
Will that stop abuse when you determine to release stock-damaging news after you had planned a scheduled sell?
heliophobicdude··on The New Clubhouse
If anyone from Clubhouse is reading this, or any entrepreneurs in this space, it's soooo hard managing whether or not I can get my friends into a new platform if some or all have conceptions that they would need an invite in order to use the app.
heliophobicdude··on A BERT for laptops, from scratch
The biggest distinction in architecture between BERT and GPT is that BERT looks both ways from a given token. This helps give context to a token. This is what made BERT great at the time because the surrounding text, before and after, could change the meaning of the token we are at. You could essentially fill in the middle, or rather correct what's in the middle after it's been said. I believe this is why Apple is using it for iOS 17's auto-correct.

GPT predicts the next word by only look back at what we have seen so far. In other words, it's auto regressive.

heliophobicdude··on Apple Sep 2023 Event: “Wonderlust”
Will the new iPhones' USB-C ports support connectivity with AR glasses? My current AR glasses work with my iPad but an iPhone would be ideal!
heliophobicdude··on Successful room temperature ambient-pressure magnetic levitation of LK-99
I thought for a second you were referring to the original post with the edit
heliophobicdude··on LK-99: The live online race for a room-temperature superconductor
I've been live following this thread:

https://twitter.com/iris_igb/status/1685731177523449856

heliophobicdude··on Argonne National Lab is attempting to replicate LK-99
Has there been any serious discussion about using GitHub for peer reviews? Specifically the pull request functionality?
heliophobicdude··on LLaMA2 Chat 70B outperformed ChatGPT
The benchmark I care about the most for my development workflow is on structured output.

Paul Gauthier made this benchmark [1] to measure correct git diffs. If you ask GPT-4 for help with your code, it can output a change in a git diff more reliably than 3.5.

My hope is that we can do that with Llama 2.

1:https://aider.chat/docs/benchmarks.html

heliophobicdude··on AWS P5 instances with H100 generally available
p5.48xlarge $98.32/hr
heliophobicdude··on IntelliJ IDEA 2023.2
This thread is about IDEA but I wish there was more awareness about the serious performance issues in CLion. I have to spend over 30 minutes for the refactor-> extract method to prompt me what I would like to name my method. Then another 30 minutes for it to execute it. It's a small c++ project too.

I appreciate the community version of IDEA being open source. I was able to look up the implementation for how java that handles this. The c++ implementation I cannot access. I've added some comments with my results from a profiled run on some old tickets about this (including one as far back as 8 years ago) but it doesn't seem to be a priority, yet!

heliophobicdude··on Will a prompt that enables GPT-4 to solve easy Sudoku puzzles be found?
The title is funny to me. We should consider a new computation complexity class for LLMs. Let's call the ones that can be solved with a prompt, Promptable. For the problems that we cannot reliably solve with a single prompt yet, let's call them non-deterministic promptable, or NP.

Question is, for most of these hard problems, is there a prompt that can solve them? Better yet, is there a prompt good enough that we collapse all of the hardest problems in NP with a single prompt?

Will we ever know if NP can be reduced to P???

heliophobicdude··on Custom instructions for ChatGPT
Seems like a system prompt for the UI version. Welcome addition!
heliophobicdude··on Ask HN: How do you find airline tickets?
I use Google Flights for most except for

SouthWest.com directly for their flights.

jsx.com for the small niche rides on a small jet

Unrelated: rome2rio.com for the whole picture including last mile to and from airports

heliophobicdude··on Ask HN: Any software to manage team rotations?
I've heard from a friend that they use GitLab's built-in management system [1]. It might not be worth paying them just for this feature though...

1:https://docs.gitlab.com/ee/operations/incident_management/on...

heliophobicdude··on Bard’s latest update: more features, languages and countries
Wired article said development will take a number of months. Perhaps we shall ask Barb?
heliophobicdude··on Bard’s latest update: more features, languages and countries
I'm really excited about Gemini. It's supposed to have some sort of planning and problem solving ability[2][3]. If it is like Tree of Thoughts [1], then I'm so hyped!

Also take a peek at which institution most of the ToT authors are at.

1: https://arxiv.org/pdf/2305.10601.pdf 2: https://blog.google/technology/ai/google-io-2023-keynote-sun... 3: https://www.wired.com/story/google-deepmind-demis-hassabis-c...

heliophobicdude··on Why Nvidia Keeps Winning
I recently found an article by Apple explaining how they optimized a transformer to work on Apple Neural Engine [1]. This was before their recent hardware refresh with 192 GB unified memory upgrade.

Apple could have a real chance to push their boxes as an alternative to dedicated GPUs.

1: https://machinelearning.apple.com/research/neural-engine-tra...

heliophobicdude··on Stability AI Launches Stable Diffusion XL 0.9
If Emad tweeted this image [1] made with SDXL, then text in image could possibly better!

1:https://twitter.com/emostaque/status/1671885525639380992

heliophobicdude··on GPT4 is 8 x 220B params = 1.7T params
Would this be before or after inference? Is there some sort of a delegation based on the matter?
heliophobicdude··on Function calling and other API updates
Not yet. I just have the GPT-4 which is 8K limit
heliophobicdude··on Apple's game porting toolkit is fantastic. Cyberpunk 2077 at Ultra on an M1 MBP
> This is the entry level M1. 15fps at 1400x900 isn't great (or even playable), but it's very impressive that it runs at all without any changes, let alone on ultra settings.

And running non-natively through Codeweavers patch and through Rosetta.

← PreviousPage 2 of 6Next →