sroussey

3 ms·

1.
by sroussey 5d ago
Why would you need multiple agents and not one agent with multiple repos?
2.
by sroussey 6d ago
There are AMD and Intel devices on similar process (not talking about A20Pro or M6 which are set to ship later this week), and they do not get the same gains. And honestly, they have historically had different markets. When the design is fo
3.
by sroussey 6d ago
HBM also trades bandwidth for latency, and your regular computing is much more sensitive to latency than bandwidth.
4.
by sroussey 6d ago
That would be like skipping land line phones for mobile...
5.
by sroussey 7d ago
what do you call bare metal in my office?
6.
by sroussey 7d ago
Curious about people’s experience here. I am working on a small model, verify by jev, and escalate to big model. Some cases, the small model is not a model but some regex. cheap-confirm-escalate Using jev as the confirm step.
7.
by sroussey 8d ago
For reference: https://huggingface.co/blog/webgpu-kernels I think it will be the basis for a rewrite of transformers.js v5, but no need for you to wait as you would likely want direct access. It is also way better than
8.
by sroussey 8d ago
Would love to see this implemented with @huggingface/kernels for shader compilation for Webgpu.
9.
by sroussey 8d ago
Those machines with GPUs still need RAM of their own, and they generally want large caches to avoid SSD penalties. You even see this spill out in the form of costs for KV cache in <1min, 5m, 1hr rates etc.
10.
by sroussey 8d ago
Someone really needed a few hundred TB to waste on inference and went looking under the rugs…
11.
by sroussey 8d ago
Having worked in hardware for a moment, everything we do in software is like this. Even C.
12.
by sroussey 8d ago
Maybe these big ai labs will uses their own devices to find and fix bugs up and down their stack and contribute that back.
13.
by sroussey 8d ago
Yeah, isn’t that Claude Codes sandbox? That drops and every npm install it taking over the world, lol.
14.
by sroussey 9d ago
Or AMD’s putting sram on a separate chip. I’m surprised they didn’t go there yet except for extra L3 instead of all of it. At some point the costs will shift the decisions.
15.
by sroussey 9d ago
Yes, exponential back off and jitter are the first things to work on, and good if you don’t have a better signal (like loss of network). Also, a simple signal status server or queue system helps to keep global state such that everyone doesn
16.
by sroussey 9d ago
So many variables, but the simple thing is to set things up like normal rate limiting (which you would want to do anyways). The one generating the errors passes back a retry time. You can add jitter here, tell low priority requests to wait
17.
by sroussey 9d ago
Yes! And maybe get a hf fused webgpu runner for that model so it’s fast!
18.
by sroussey 10d ago
The hybrid nature of this thing is not a detriment, nor does it make it an LLM.
19.
by sroussey 11d ago
I dunno, I would consider Waymo and Tesla to have frontier models. I think AlphaFold and related are also frontier models. Being an LLM does not seem like the qualifier for frontier.
20.
by sroussey 12d ago
I have the same problem when i touch a result and it changes the result line at the exact moment.
21.
by sroussey 12d ago
And there was the Tesla thing CNAMEing time server pools and hiring people to pen test, which sent automated attack systems on volunteers servers. Last I heard, Tesla et al didn't even care enough to respond.
22.
by sroussey 12d ago
Should this not all happen before the PR is created?
23.
by sroussey 13d ago
All my photos are HEIC. Why not just use what I have and not translate?
24.
by sroussey 13d ago
How is your location data protected? They may not be able to force a company to hand it over without a warrant… but they can just buy it.
25.
by sroussey 13d ago
Give Tesla the wrong time.
26.
by sroussey 13d ago
What models can it run?
27.
by sroussey 13d ago
Hugging face is working on something like this where well known models get fused into a single implementation.
28.
by sroussey 13d ago
Language models have always had an issue with negatives. A negative like do “not” xyz is just not encoded the same as spelling out what you want vs what you don’t want. Harder to write though.
29.
by sroussey 14d ago
> Private surveillance that immediately sends any and all crimes to law enforcement is indistinguishable from government surveillance and should treated the same way. Don’t use Snapchat then… https://www.gadgetreview.com/
30.
by sroussey 14d ago
Dictation, text to speech, some of that computational photography, etc too!

More ›