karolist

Fair point, thanks for clarification, it seems this was first proposed in https://arxiv.org/pdf/2405.04434? I was confused by your title mentioning DeepSeek but then first paragraph revert to "...language models like ChatGPT and DeepSeek faster at generating text".

What's specific to deepseek here that other models do not use, or are you just riding the keyword wave?

It does that all the time even now. Play with temp, top_k, min_p and nothing is preventing you from getting combinations and order of tokens previously not encountered in history.

I'm not sure what's going on with TabbyAPI's github metrics, but exl2 quants are very popular among nvidia local LLM crowd and TabbyAPI comes in tons of reddit posts of people using it. Might be just my bubble, not saying they're not accurate, just generally surprised such a useful project has under 1k stars. On the flip side, LLMs will hallucinate about TabbyML if you ask it TabbyAPI related questions, so I'd agree the naming is unfortunate.

Techcrunch didn't omit the lawsuit information which alleges Nvidia farmed YouTube to train this, I found this more useful than pure marketing spin on nvidia.com.

You don't know enough to know the full picture, easy to draw conclusions sure, but they are most likely incorrect.

llama.cpp can run LLMs on CPU. iGPU can also use system memory, the novel thing is not that, it's that the LLM inference is mostly memory bandwidth bound and memory bandwidth of a custom built PC with really fast DDR5 RAM is around 100GB/s, nVidia consumer GPUs at the top end are around 1TB/s, with mid range GPUs at around half that. M1 Max has 400GB/s, M1 Ultra is 800GB/s, but you can have Apple Silicon Macs with up to 192GB of 800GB/s memory usable by GPU, this means much faster inference than just CPU+system memory due to bandwidth and more affordable than building a multi-GPU system to match the memory amount.

In case someone wonders, AMD has its own equivalent - https://en.m.wikipedia.org/wiki/AMD_Platform_Security_Proces...

Crowley saw demons as manifestations of the mind's architecture - real in their effects, but originating from within rather than existing as independent external entities. To that effect, they are very real, in the same way that the color green is real - subjectively experienced by humans with normal color vision, yet meaningless to a being with a different sensory architecture.

Majority of learning surface will come from framework and not the language, if you need Python elsewhere just learn that as well, but this shouldn't move framework choice much.

Recommendation engines work on vast amounts of data they have on you and whatever made you speak about thing X was likely preceded by your internet activity which is not very unique as a precursor to speaking about X. In other words, if other people do Y on the internet and then end up doing stuff related to X, the recommendation engine will show you X just because you also did Y.

The other explanation is one of your contacts who were part of the conversation did things that either directly related to thing X, which you spoke about, or something the algorithm see other people do that relates to X, and you got shown ads based on your affiliation to this person.

I've also worked at FAANG and never seen proof to such claims anywhere in the code, and with the amount of people working there who care about these issues deeply I'd expect this to leak by now, if this happens but is siloed...

Eco guidelines apply to manufacturer's fleet sales as a whole, meaning a larger manufacturer which makes EVs and eco friendly ICE variants has the "budget" to offer models that are not eco friendly, such as your examples. Suzuki doesn't have offers on the market that would offset the Jimny's pollution so they can't sell it anymore.

I feel so to. The cloud billing model shifts responsibility to users under the guise of "flexibility" and "customization". Imagine a car rental company that charges you by the millisecond for every component - engine camshaft revolutions, tire rotations, windshield wiper activations, seat heating time and so on, but has the ability to set up alerts for each component so that customers "control" their usage and budget. It's just user hostile, risky by default cost model.

Huge congratulations team, you were always one of the few Lithuanian companies I'm proud to talk about.

Didžiausi sveikinimai!

Bought the same machine from the same store last December, been using it ever since with a big grin on my face. I'll probably consider upgrading in 2028 or later.

Looks like lametric alternative?

Yes, apologies, can't edit now

Roughly, at Q8 the model sizes translate to GB, so ~3 and ~70GB.

Felt like talking to a person, I couldn't bring myself to treat it like a piece of code, that's how real it felt. I wanted to be polite and diplomatic, caught myself thinking about "how I look to this person". This brought me thinking of the conscious effort we put in when we talk with people and how sloppy and relaxed we can be when interacting with algorithms.

For a little example, when searching Google I default to a minimal set of keywords required to get the result, instead of typing full sentences. I'm sort of afraid this technology will train people to behave like that when video chatting with virtual assistants and that attitude will bleed in real life interactions in societies.

Sage/Breville Dual Boiler and do the slayer mod + drip tray mod, you can then pull espresso that rivals any machine out there for a fraction of the cost. https://www.youtube.com/watch?v=JmQgxQ5Higw

The machine is old enough to be well understood with documented mods and fixes, it's dual boiler, triple PID, extreme temp stability, fill from the front, drip tray indicator. I love it.

https://home-barista.com/espresso-machines/breville-dual-boi...

I'd argue most people don't even like coffee, they drink it for the caffeine.

SQLite is a library, the process still exists, it's just that it's part of your app now.

I think the analogy works well but you must understand the reason for flywheels, at least in cars. A flywheel in a manual transmission car acts as an energy storage and output smoothing device. An internal combustion engine delivers non-continuous power impulses, as explosion happens in the cylinder, the force pushes the piston down, which in turn pushes on the crankshaft which results in rotational power impulse. But these impulses happen separately which results in gaps where there is no force applied to the crankshaft. Without a flywheel, your car would vibrate badly, flywheel captures these impulses and delivers continuous output.

Diesel manual transmission cars have heavier flywheels because higher compression in the cylinder results in higher torque applied to the crankshaft, therefore you need more mass to store more energy. Heavier flywheel has more rotational inertia, thus this makes revving a diesel engine seem lethargic compared to a petrol car, the heavier flywheel takes more time to reach desired RPM, thus revving a car with a heavy flywheel results in slow RPM pick up.

A fairly recent trend is putting a bit heavier flywheel to even small petrol cars, this makes them harder to stall at the expense of driver's engagement, this combined with custom fuelling maps make new petrol cars really easy to drive but not really engaging. It is common to put a lighter flywheel to cars being raced, you get faster revs and better engine response at the expense of the engine being easier to stall.

Automatic transmission cars, at least those with a torque converter, also have a flywheel, but since the design of torque converter allows for non-direct engine-transmission coupling, most of the smoothing can happen inside the torque converter itself, thus the flywheel is extremely light.

It's funny how transparent LLM writing is, I hope it stays that way.

A transfer case is something BMW calls the thing that attaches to the side of your transmission to provide power to front wheel shafts, this does not mean the car is 4WD. Yes, still confusing.

https://awd.tech/de/apps/catalog/transmission/atc300-bmw

I think so too. Also, what's the point of using git with commit messages like that? https://github.com/Dicklesworthstone/visual_astar_python/com...

LoL.

Intrigued by your Workflowy mention. Any resources where I could learn your style of workflow with it?

I think the problem is due to people coming to Vim after they've used an IDE and want an IDE like experience, that's how you end up with Lua module tracelogs when you open nvim and spend time fighting the tools rather than using them. I too went though that phase but in the end I realized that I actually just enjoy vim motions, if I have them in my IDE it's already good. I'm back to Intellij with ideavim plugin and it's been great so far. I know things like lazyvim made the setup more manageable but I've never tried it.

The amount of YouTube tech grifters support this claim. Any seriously skilled dev will barely have an audience but channels skimming the water grow to millions.

This article is too basic and does not introduce anything you'd encounter in a typical Go project. If you want introduction to Go testing I recommend just reading the official docs https://pkg.go.dev/testing and understanding how to write table driven tests - https://go.dev/wiki/TableDrivenTests.

Going beyond what's built in, get familiar with https://github.com/stretchr/testify as that's used a lot.