X-Ryl669 · HackerTrans

hace 16 díasdiscuss

I'm not 100% sure it's not possible. If (I don't know) it's possible to freeze the temperature of the model so it's deterministic, and if you could make a map of produced words back to tokens (via HMM probably), then you can probably alter a minimal input and observe the output to model it. If you perform waves of such minimal alterations, you can expect to be able to locate the distance where each alteration impact the model (the idea being that a small alteration on output is likely due to the last layers of the models, and a small alteration is likely due to the deeper layer). Once you've located most of the last layer(s?) weights, you can try to solve for them. With a hundreds of billions weights model, the last layers will likely be so huge that it's probably unfeasible technically, but it's theoretically possible.

X-Ryl669

el mes pasadodiscuss

No it doesn't. Think of the intro of We'll rock you and show me how you perform that on your piano.

X-Ryl669

hace 3 mesesdiscuss

That's true. And, in my understanding, what OO did, falls exactly in paragraph b and d. The license doesn't describe what is the "Appropriate Legal Notices" and OO provided a description for it: its logo and its trademark.

X-Ryl669

hace 3 mesesdiscuss

The license contains (section 7):

> [you may] supplement the terms of this License with terms: > >[...] > > b) Requiring preservation of specified reasonable legal notices or > author attributions in that material or in the Appropriate Legal > Notices displayed by works containing it; or > > c) Prohibiting misrepresentation of the origin of that material, or > requiring that modified versions of such material be marked in > reasonable ways as different from the original version; or > > d) Limiting the use for publicity purposes of names of licensors or > authors of the material; or > > e) Declining to grant rights under trademark law for use of some > trade names, trademarks, or service marks;

So the requirement of branding and attribution aren't "further restriction" (which, in this context, means a restriction that is not in the AGPLv3 license text). It's after section 7's list of allowed restrictions, which, paragraph b, contains "require preservation of [...] legal notices or [...] attributions", paragraph d is made to prevent misuse of the original author reputation, and paragraph e to prevent misuse of trademarks, so they, IMHO, are all legitimate.

X-Ryl669

hace 3 mesesdiscuss

This is bullshit. Read section 7 paragraph b of the AGPLv3 license, and you'll see that OnlyOffice did what they were allowed to do. The Nextcloud blog here is spreading FUD on their partner which will likely cause more damage to them if OO's lawyer starts to have a look on this. People don't read license text, but they should, because they think AGPL is like GPL which is not the case, they are additional restrictions to the former.

X-Ryl669

hace 3 mesesdiscuss

I think you're claiming wrong stuff here. AGPLv3 section 7 paragraph b) expressively authorize the author to require an attribution in the derived work or copy. What Nextcloud did was to remove this attribution, so they actually mooted their own right to use the code under that license. There's nothing related to trademark or branding violation here. If OnlyOffice attacked Nextcloud for using their TM or brand for respecting the license, they would be debunked at a trial (if it even reach a trial), since they expressively allowed the use of the attribution in distributing their work with this license. Note: This license doesn't give you the right to use the branding of OnlyOffice on a derived product and claim it's yours or you're acting as them, that's a complete different usage case here.

X-Ryl669

hace 3 mesesdiscuss

I think you're confused by the term "permissions". You can give more freedom to the license and a copier can remove them as long as it doesn't remove the freedom that are in AGPLv3. The OnlyOffice team claim comes from the next paragraph of section 7:

> Notwithstanding any other provision of this License, for material you add to a covered work, you may [...] supplement the terms of this License with terms:

> b) Requiring preservation of specified reasonable legal notices or author attributions in that material or in the Appropriate Legal Notices displayed by works containing it; or c) Prohibiting misrepresentation of the origin of that material, or requiring that modified versions of such material be marked in reasonable ways as different from the original version; or

This is what they did and what the other part stripped from their blatant copy. So no, removing the logo or the OnlyOffice terms therefore seems forbidden by the license itself, revoking it for the other part, thus they are now making a counterfeit.

X-Ryl669

hace 4 mesesdiscuss

Since there's a single provider of this "specification", which is more or less Google, why even bother calling it a standard? Even more that Google's implementation is not following the published standard, breaking potential competing implementation in so many subtle ways.

X-Ryl669

hace 5 mesesdiscuss

No you can't. dlopen signature takes a file path, not a memory range. And if you start to save the libraries to the filesystem before opening them, there's no difference to shipping an archive directly and skip the trouble of your own archive code.

X-Ryl669

hace 6 mesesdiscuss

Read that: https://blog.cyril.by/fr/software/an-expressive-text-to-spee... and you'll find answers to your remarks

X-Ryl669

hace 8 mesesdiscuss

I don't get why you process the whole connector board at once. If I understand correctly, you're connecting individual & identic boards to your board. So each connector on you giant board is actually dealing a bunch of small board right?

In that case, can't you exploit the inherent symmetry in the design here to only route a quarter of your connectors and then mirror/rotate the result for the other one? Or, if you have a X*X matrix, route one size minus the corners and replicate to the other sides?

Also, with such a huge connection board, it smells a NIH issue here. I think you'd better serialize the IO to a bus (whatever) and few lines and perform the connection in software (in a GoWin FPGA for example, both extremely cheap and quite powerful). Just think of the harness you'll need to build to fit the connectors in. The obvious routing bugs, and so on. Any maintenance will be a nightmare, if you need to swap 2 pins on a connector or re-run the routing.

X-Ryl669

hace 9 mesesdiscuss

No you can't since the compiler will likely inline the syscall (or vsyscall) in your functions. So there's no way to know the instruction pointer is in the right section. The only way is to pay for no-inline cost and have a wrapper that's calling the syscall, so it's a huge cost to pay for a very rare feature (cancelling a thread abruptly is a no-no in most coding conventions).

X-Ryl669

hace 10 mesesdiscuss

I'm using https://glaxnimate.org/ . It's not well known but it's stable and works well for all my needs.

comments