treesciencebot

7 ms·

61.
TripoSR: Fast 3D Object Generation from Single Images(stability.ai)
5 points by treesciencebot 3y ago|0 comments
62.
by treesciencebot 3y ago
> Above I alluded to the fact that we briefly ran ephemeral, interactive, session-lived processes on Kubernetes. We quickly realized that Kubernetes is designed for robustness and modularity over container start times. Is there a clear e
63.
by treesciencebot 3y ago
It allows for research to continue, which might eventually benefit everyone. The primary advantage in my mind is giving academy a chance to learn from it and community to build cool stuff on top of it.
64.
by treesciencebot 3y ago
> (1) They can give away the model but sell an API - but they can’t serve a model as cheap as Goog/Msft/Amzn who have better unit economics on their cloud and better pricing on GPUs (plus custom inference chips). Which has a si
65.
Qualcomm's Adreno 530, a Small Mobile iGPU(chipsandcheese.com)
2 points by treesciencebot 3y ago|0 comments
66.
by treesciencebot 3y ago
pixel art is a particularly hard thing for these models to do, especially without further fine tunes or loras. but I'm pretty sure you should be able to get that quality with one of nerijs's loras [0]. But for now, i'd do som
67.
by treesciencebot 3y ago
Spatial prompt adherence is a general missing piece is SDXL (or previous versions of the SD). Hoping that the SD will get it into a good shape as your examples! Test the example on Stable Cascade as well (latest open-weight stability model)
68.
by treesciencebot 3y ago
Yep, this is using SDXL Lightning underneath which is trained by ByteDance on top of Stable Diffusion XL and released as an open source model. In addition to that, it is using our inference engine and real-time infrastructure to provide a s
69.
Show HN: Real-time image generation with SDXL Lightning(fastsdxl.ai)
444 points by treesciencebot 3y ago|104 comments
70.
Accelerators(github.com)
2 points by treesciencebot 3y ago|0 comments
71.
Jasper Expands by Acquiring Image Platform Clipdrop from Stability AI(jasper.ai)
1 points by treesciencebot 3y ago|0 comments
72.
by treesciencebot 3y ago
Quite nice to see diffusion transformers [0] becoming the next dominant architecture on the generative media. [0]: https://twitter.com/EMostaque/status/1760660709308846135
73.
Real-time text-to-image generation powered by SDXL Lightning(fastsdxl.ai)
5 points by treesciencebot 3y ago|0 comments
74.
Stable Diffusion XL Lightning(huggingface.co)
3 points by treesciencebot 3y ago|0 comments
75.
by treesciencebot 3y ago
per-chip compute is not the main thing this chip innovates for fast inference, it is the extremely fast memory bandwith. when you do that, you'll loose all of that and will be much worse off than any off the shelf accelerators.
76.
by treesciencebot 3y ago
there are providers out there offering for $0 per million tokens, that doesn't mean it is sustainable and won't disappear as soon as the VC well runs dry. Am not saying this is the case for Groq, but in general you probably should
77.
by treesciencebot 3y ago
The main problem with the Groq LPUs is, they don't have any HBM on them at all. Just a miniscule (230 MiB) [0] amount of ultra-fast SRAM (20x faster than HBM3, just to be clear). Which means you need ~256 LPUs (4 full server racks of c
78.
by treesciencebot 3y ago
Yep, we are currently in private beta for custom models. Hit us at hello@fal.ai for access!
79.
by treesciencebot 3y ago
Just as a top-level disclaimer, I'm working at one of the companies in "this" space (serverless GPU compute) so take anything I say with a grain of salt. This is one of the things we (at https://fal.ai ) working ve
80.
by treesciencebot 3y ago
The examples are most certainly cherry-picked. But the problem is there are 50 of them. And even if you gave me 24 hour full access to SVD1.1/Pika/Runway (anything out there that I can use), I won't be able to get 5 examples
81.
by treesciencebot 3y ago
If we go from DALL-E 3, it won't be nowhere near competitive while they have the superior ground. Generating a high quality 1024x1024 image with costs around ~$0.002, but $0.08 on DALL-E 3 (20x more expensive per-image). For videos wit
82.
by treesciencebot 3y ago
This is leaps and bounds beyond anything out there, including both public models like SVD 1.1 and Pika Labs' / Runway's models. Incredible.
83.
by treesciencebot 3y ago
Just to correct the record, both $1.15 per A100 and $2.24 per H100 require a 3-year-commitment. On-demand prices are 2.5X that.
84.
by treesciencebot 3y ago
in my observation, it yields amazing perf at higher batch sizes (4 or better 8). i assume it is due to memory bandwith and the constrained latent space helping.
85.
by treesciencebot 3y ago
Uh, thanks for noticing it! We generally turn it off for popular models so people can see the underlying inference speed and the results but we forgot about it for this one, it should now be auth-less with a stricter rate limit just like ot
86.
by treesciencebot 3y ago
I think the model architecture (training code etc.) itself is still under MIT while the weights (which was the result of training in a huge GPU cluster as well as the dataset they have used [not sure if they publicly talked about it] is und
87.
Examining AMD's RDNA 4 Changes in LLVM(chipsandcheese.com)
4 points by treesciencebot 3y ago|0 comments
88.
by treesciencebot 3y ago
with MLO and the promised latency decreases, even at 300Mbps, it should significantly make a difference on how "snappy" everything is (and how reliable video calls are).
89.
Qualcomm's Adreno 530, a Small Mobile iGPU(chipsandcheese.com)
14 points by treesciencebot 3y ago|3 comments
90.
by treesciencebot 3y ago
I don't understand how people find 3V at 180mA usable, isn't it like 0.5 watts?

More ›