So an encoder-only model with a classifier trained on the heads or something? DeepSeek recently switched to an encoder-decoder architecture in an attempt to get the best of both worlds (fast prefill while preserving generation capability), I wonder if that might be the future?
>This spills over into other processes wanting to use the GPU, namely the WindowServer.
Why does this spill over? Unlike CPU which is multiplexed by the kernel's scheduler (so infinite loops can't lock out other programs), is the GPU not multiplexed in the same fashion?
Often not in the same way - even if there are multiple queues (that can be given a priority), they're often limited in what they can schedule between.
Often there's shared resources that are statically allocated to shaders (register space, local memory etc.) that means you often can't "just" add a new task if those shared resources are already in use. But not using those resources to their full would cause performance issues.
And the internal state of a GPU is often very large, much larger than a CPU, so suspending the current tasks, saving out their state and replace it with a "higher priotity" one can be very expensive - so often an afterthought of support at best.
Otherwise GPUs typically do context "pre-emption" by basically being cooperative and just injecting yield statements in the command queue or on things like tile boundaries for tile based renderers. So the smallest chunk of work they can yield between ends up actually being quite large, and with a full user-supplied program in the middle
you can get another LLM to verify / if the lean doesn't have `sorry` used to skip certain parts of the proof etc. It's much easier once it's in lean4 because checks like that can be done computationally.
Given that we have self-driving cars, isn't this easier if someone really wanted? I guess compared to cars the marginal savings is not worth it though.
I've never quite understood the hate either. But the way I see it is that it's not fundamentally inferior to explorer, but rather just has annoyances and warts that Apple has bizarrely not fixed for decades which makes tedious to use for anything other than casual browsing.
You can still navigate by path via cmd+shift+g (this works in the file picker too), but I feel like the popup animation used to be a lot quicker on older versions. You can customize views, but to get them to actually stick you have to use an arcane incantation (delete all existing ds_store, then go to view options menu and hit "make default").
You can cut-paste but you have to use cmd+c and cmd+option+v instead of cmd+x. There is an option to merge, it only appears under conditions: only when copy/pasting and only when "alt" is held in the dialog I think.
It also is really bad at browsing networked volumes, partly due to poor defaults as it tries to litter ds_store and generate thumbnails.
Opening a new tab always takes you back to the root directory instead of cloning your existing view. All of it is stuff that could be added in a simple point update rather than a ground-up rewrite.
Yeah, can't read anything on xitter due to the gigantic dickover they put over the content when you don't have an account. A screenshot would've been more helpful than an xitter link.
reply