HackerTrans
TopNewTrendsCommentsPastAskShowJobs

logotype

no profile record

Submissions

Speculative pre-positioning: off-path decode for stateful inference sessions

arxiv.org
1 points·by logotype·قبل 8 أيام·1 comments

Stateful Inference for Low-Latency Multi-Agent Tool Calling

arxiv.org
2 points·by logotype·الشهر الماضي·1 comments

Attention Once Is All You Need: Stateful Transformers

arxiv.org
3 points·by logotype·قبل شهرين·4 comments

[untitled]

1 points·by logotype·قبل 3 أشهر·0 comments

[untitled]

1 points·by logotype·قبل 3 أشهر·0 comments

New inference engine faster than vLLM, SGLang, TRT-LLM

layerscale.ai
2 points·by logotype·قبل 5 أشهر·3 comments

[untitled]

1 points·by logotype·قبل 5 أشهر·0 comments

[untitled]

1 points·by logotype·قبل 6 أشهر·0 comments

Ask HN: Need ArXiv endorsement for LLM inference paper

1 points·by logotype·قبل 6 أشهر·1 comments

Ask HN: How to get funding for your product?

3 points·by logotype·قبل 8 أشهر·2 comments

Nobody Talks about Failure

3 points·by logotype·قبل 9 أشهر·0 comments

comments

logotype
·قبل 8 أيام·discuss
With native support for sessions in an inference engine we can make use of idle GPUs... doing work!
logotype
·الشهر الماضي·discuss
I bought a server with 768GB RAM a couple of years ago, for cheap. Today just the RAM itself is worth more than I paid for the entire server. DDR4... Same thing with SSD, two years ago you could buy 4TB SSDs for like 250 USD, and today its more than $700. Madness!
logotype
·الشهر الماضي·discuss
Fastest tool calling on the planet!
logotype
·الشهر الماضي·discuss
Pin-drop sweepstakes
logotype
·الشهر الماضي·discuss
Attended Dreamhack in the late 90s. Fun times! I was always AMAZED by the demoscene.
logotype
·قبل شهرين·discuss
Tumbleweed rolls across the timeline
logotype
·قبل شهرين·discuss
The problem is that 95% of the products on Amazon are just chinese garbage. Most people do not research the products they are buying, they are buying because of the convenience. The "default" quality of products on Amazon are utter complete overpriced shit. Even if you have done your research and ACTIVELY look for a specific locally manufactured product on Amazon you're flooded with sponsored products with unpronounceable names. There's a whole hidden industry about Chinese companies have a streamlined process to create a shell company, apply for various trademarks to skirt rules even if their product is obviously copied. It's horrible if you look into it. I try really hard to only support companies that manufactures products in the US/EU/UK. You should do, too! It's good for the economy.
logotype
·قبل شهرين·discuss
crickets
logotype
·قبل شهرين·discuss
reads like Salesforce to me, ugh! Enterprises are paying so much to blatantly vendor-lock in themselves using hundreds of "Salesforce engineers". It's baffling to me.
logotype
·قبل شهرين·discuss
Happy to answer any questions you might have.
logotype
·قبل شهرين·discuss
Deeply nostalgic! Thanks for sharing.
logotype
·قبل 3 أشهر·discuss
It’s a cost I’m willing to pay. You’re paying for a fully domestic supply chain. The cotton is grown in the U.S., then cleaned, spun into yarn, made into fabric, dyed, and finally cut and sewn in Los Angeles. Every step happens in the United States, supporting local farming, manufacturing, and labor.
logotype
·قبل 3 أشهر·discuss
I meant American Giant jeans https://www.american-giant.com/collections/mens-bottoms :)
logotype
·قبل 3 أشهر·discuss
You can notice the quality difference when things are actually built to last. I work a lot with electronics and need good work-holding tools. I didn't use to focus on that and just bought cheap crap. I then heard of PanaVise (made in Reno, NV) and got their tools, absolutely SOLID. Over-built, heavy and built to LAST!
logotype
·قبل 3 أشهر·discuss
Same here.. i bulk buy tons of AG (American Giant) t-shirts (Premium Slub Crew Tee is awesome) and tons of their jeans too.
logotype
·قبل 3 أشهر·discuss
Support brands with values and local manufacturing. For example: American Giant, Origin, Crye Precision, Randolph Engineering, American Optical, and many more.
logotype
·قبل 3 أشهر·discuss
There’s a lot of wasted compute with stateless inference. We set out to solve that with a new computational model for transformers, and only process the delta between requests. That’s how we can achieve crazy low latency tool calling with LayerScale. Check it out https://layerscale.ai and technical whitepapers out next month!
logotype
·قبل 3 أشهر·discuss
Try it out and see how fast inference can be for agentic workflows. Works on CUDA and ROCm. Feedback appreciated!
logotype
·قبل 4 أشهر·discuss
I can access it from the UK
logotype
·قبل 4 أشهر·discuss
Just a OFCOM basic business radio license (UK)