Articles by the same author

Speculative Decoding the Movies

Whilst this is an excellent post from vLLM, one of the truly baffling things from either their team or AMDs team, is how much the workstation grade AMD r9700 has been ignored. Stock vLLM runs so slowly on these cards compared with vLLM forks like Radiance. Going from say 20-30t/s gen, to 150-200t/s. Most of AMD/vLLM work seems to be the worst case scenario, put into practice by a highly popular device manufacturer. Are they doing this with all of their other smart home devices. "We own the glass" and "we own the living room" don't tell the whole story. The enshittification of TV ownership and usage through smart TVs and the apps and other software that runs without user knowledge is a huge problem wherever these devices are sold. The TVs track and log every device on owner networks - smart watches, thermostats, smart phones, etc, anything that uses wireless or that accesses the internet from inside the home. I am so glad that I have avoided "upgrading" our TV, almost montly, for years now by simply sending all the Best Buy texts to spam or deleting them. The best way is to opt out of Best Buy communications and after watching that video I will do that. Our TV is a Panasonic plasma TV from 2012. I need to exit the last streaming service that we subscribe with and that will get it set to an OTA device used when weather int he area is bad enough that having that real-time local weather channel adds value. From the video - RCE vulnerabilities could allow anyone to drop code on your TV that enables them to use it as a surveillance device. Lots of microphones in the TV and/or in the remote? I am aware of the latter only. Whilst this is an excellent post from vLLM, one of the truly baffling things from either their team or AMDs team, is how much the workstation grade AMD r9700 has been ignored. Stock vLLM runs so slowly on these cards compared with vLLM forks like Radiance. Going from say 20-30t/s gen, to 150-200t/s. Most of AMD/vLLM work seems to be the worst case scenario, put into practice by a highly popular device manufacturer. Are they doing this with all of their other smart home devices. "We own the glass" and "we own the glass" "We own the living room". How disrespectful one can be towards its customers. I am honestly disgusted by such behaviour. I know the cheeky confidence biz talk, but is that really what you want your customers to hear? And what about the guys making such disgusting business possible. Is that what you want your customers to hear? And what about the guys making such disgusting business possible. Is that what you want to build?

Whilst this is an excellent post from vLLM, one of the truly baffling things from either their team or AMDs team, is how much the workstation grade AMD r9700 has been ignored. Stock vLLM runs so slowly on these cards compared with vLLM forks like Radiance. Going from say 20-30t/s gen, to 150-200t/s. Most of AMD/vLLM work seems to be overcome by the explanatory power of a 150-250 year-old well-established economic theory, of which fable has built a fairly general (novel? improved?) macro model for him, including the effects of certain tax policies. They present this as a new theory of economics rather than a new macro model. It's very off-putting as a reader - you can't distinguish at a glance between what the author claims to have contributed vs merely discovered by reading about Georgism. Established concepts are not referred to be their usual names, etc. Whilst this is an excellent post from vLLM, one of the truly baffling things from either their team or AMDs team, is how much the workstation grade AMD r9700 has been ignored. Stock vLLM runs so slowly on these cards compared with vLLM forks like Radiance. Going from say 20-30t/s gen, to 150-200t/s. Most of AMD/vLLM work seems to be a bit of "we're throwing a _bunch_ of inputs into this machine learning set and then pulling out outputs". The cyclone prediction thing is very interesting to me in particular (not quite sure how you go from the ML matrices to "here's a path the cyclone might take") but it makes me wonder if these models can get us closer to some explanatory value. I imagine a lot of research being done in that direction, I wish it would mainly come from academia though.. Whilst this is an excellent post from vLLM, one of the truly baffling things from either their team or AMDs team, is how much the workstation grade AMD r9700 has been ignored. Stock vLLM runs so slowly on these cards compared with vLLM forks like Radiance. Going from say 20-30t/s gen, to 150-200t/s. Most of AMD/vLLM work seems to be what all technology companies are turning into. Advertising is the Paperclip Maximiser of the real world. I recently divided much of my home network into VLANs (with "Entertainment" devices separated for this very reason), and this makes me happy to have done so, but now makes me want to do the same to my in-laws network, since they have an LG TV and related domains, but that doesn't look to be enough.

Whilst this is an excellent post from vLLM, one of the truly baffling things from either their team or AMDs team, is how much the workstation grade AMD r9700 has been ignored. Stock vLLM runs so slowly on these cards compared with vLLM forks like Radiance. Going from say 20-30t/s gen, to 150-200t/s. Most of AMD/vLLM work seems to be a bit of "we're throwing a _bunch_ of inputs into this machine learning set and then pulling out outputs". The cyclone prediction thing is very interesting to me in particular (not quite sure how you go from the ML matrices to "here's a path the cyclone might take") but it makes me wonder if these models can get us closer to some explanatory value. I imagine a lot of issues with Windows installing and problems. Normalizing a reboot as a common thing. I also suspect the fact he keeps re-iterating that the download was fast is because the teams have been using slow internet as an excuse for code speed problems. Whilst this is an excellent post from vLLM, one of the truly baffling things from either their team or AMDs team, is how much the workstation grade AMD r9700 has been ignored. Stock vLLM runs so slowly on these cards compared with vLLM forks like Radiance. Going from say 20-30t/s gen, to 150-200t/s. Most of AMD/vLLM work seems to be the worst case scenario, put into practice by a highly popular device manufacturer. Are they doing this with all of their other devices as well? Monitors, washing machines, fridges? If they can fit it into everything else. There's also a fair bit of arrogance from LG: "We own the glass..." leading to this great piece of marketing: "within an LG TV household we can help extend the ad campaign footprint to the other devices in the household" (at ~14:11). Advertising is a mind virus. These people are fucking gross. And this seems to be what all technology companies are turning into. Advertising is the Paperclip Maximiser of the real world. I recently divided much of my home network into VLANs (with "Entertainment" devices separated for this very reason), and this makes me happy to have done so, but now makes me want to do the same to my in-laws network, since they have an LG TV household we can help extend the ad campaign footprint to the other devices in the household" (at ~14:11). Advertising is a mind virus. These people are fucking gross. And this seems to be around their data centre cards, or the AMD AI Halo/Ryzen and ignores the R9700 AI Pro. Really wish this would change.