Is Allie a 'Leela net player'?

[Alayan]

So what I guess what I'm saying is, including Allie is more fun than having two engines compete for two SuFi spots? And I'm not sure there's any other question to ask.

If you think having close derivatives doesn't affect fun and rather increase it, I guess you'll be happy when all the original non-SF non-Leela engines of relevant strength go out of development while you'll get 10 NNUE players to compete with Leela (but still lose to Stockfish).

[chrisw]

Chess engine programmers really should have a pile of sand in the backyard plus a kiln and make their own CPUs

What is this even supposed to mean?

You don't get metaphor?

[Tony P.]

^^ SlowBlitz, RofChade, Schooner, and Pedone (and Komodo, though I'm not sure how long it will stick around) are closed-source, and I trust that their authors are honest enough not to copypaste GPLed code. Also, Winter is on a very original route, and I haven't stopped believing in it yet.

[the_real_greco]

So what I guess what I'm saying is, including Allie is more fun than having two engines compete for two SuFi spots? And I'm not sure there's any other question to ask.

If you think having close derivatives doesn't affect fun and rather increase it, I guess you'll be happy when all the original non-SF non-Leela engines of relevant strength go out of development while you'll get 10 NNUE players to compete with Leela (but still lose to Stockfish).

I think you misunderstood me. I didn't generalize to "close derivatives". I made a claim about the effect of Allie in this specific circumstance.

[the_real_greco]

Then why have uniqueness rules in the first place? Why doesn't TCEC do what CCCC does and say, 'here are the engines that are playing this season and we as the organisers have the right to choose which engines get to participate in the tournament and which engines do not, to ensure maximum fun in the tournament'.

You're preaching to the choir! But I will say, the difficulty lies in who you're letting pick the engines. Being able to reference rules gives the veneer of objectivity.

[Madeleine Birchfield]

Moving the conversation back into this thread:

The discord handle of the NVIDIA developer who has contributed CUDA code to both Allie and LC0 is ankan. I assume that’s what you are referring to as “leela backend?” To top it all off, the code isn’t even chess specific.

So should the CUDA backend be considered as a library instead of Leela specific code, if it is more general than Leela? Should NNUE be considered a library instead of Stockfish specific code as well then? NNUE isn't chess specific either; it originated in shogi. Should the rules governing the CUDA backend library therefore be the same as the rules governing NNUE code in TCEC, and should the rules governing both be more restrictive regarding engine uniqueness or less restrictive?

[Madeleine Birchfield]

You're preaching to the choir! But I will say, the difficulty lies in who you're letting pick the engines. Being able to reference rules gives the veneer of objectivity.

TCEC already suffers from perceived bias, in opening choices and hardware choices, along with engine participation. Especially with hardware, there seems not to be any inherent objectivity in TCEC's choices in the hardware, the hardware used is what TCEC says it will be. I don't see much difference with choosing engines then, if there is already precedent in hardware choices. No need for engine rules, make it invite only. And make it clear that each engine has to be re-invited to the tournament each season to participate that season.

[dkappe]

Moving the conversation back into this thread:

The discord handle of the NVIDIA developer who has contributed CUDA code to both Allie and LC0 is ankan. I assume that’s what you are referring to as “leela backend?” To top it all off, the code isn’t even chess specific.

So should the CUDA backend be considered as a library instead of Leela specific code, if it is more general than Leela? Should NNUE be considered a library instead of Stockfish specific code as well then? NNUE isn't chess specific either; it originated in shogi. Should the rules governing the CUDA backend library therefore be the same as the rules governing NNUE code in TCEC, and should the rules governing both be more restrictive regarding engine uniqueness or less restrictive?

Don’t evade the point. You are fluttering your hands and referring to some “leela backend” that Allie is using. Not true. What there is instead is some highly optimized linear algebra CUDA code written by one individual and contributed to both projects.

Stop making yourself look ridiculous.

[Daniel Shawul]

Moving the conversation back into this thread:

The discord handle of the NVIDIA developer who has contributed CUDA code to both Allie and LC0 is ankan. I assume that’s what you are referring to as “leela backend?” To top it all off, the code isn’t even chess specific.

So should the CUDA backend be considered as a library instead of Leela specific code, if it is more general than Leela? Should NNUE be considered a library instead of Stockfish specific code as well then? NNUE isn't chess specific either; it originated in shogi. Should the rules governing the CUDA backend library therefore be the same as the rules governing NNUE code in TCEC, and should the rules governing both be more restrictive regarding engine uniqueness or less restrictive?

Yes, NNUE or other NN probing code is something that fits the description of a library nicely.
Re-inventing the wheel is not a mark of originality despite what some say here.
For example, I use tensorflow library for probing any neural network on the CPU, and TensorRT for the GPU (I could probably
skip this one and use tensorflow for the GPU too). Tensorflow can do CuDNN or TensoRT or OpenCL etc, so in this regard the gazillion
leela backends are a waste of time. Infact, it is my opinion that this has proven detrimental to experimentation with different net architectures.

Even for NNUE you can probably use Tensorflow for the dense layers and use INT8 optimization to get a similar level of accelration.
While the first layer can be calculated by the engine from weights retreived from the tensorflow graph.
The current NNUE training code is adhoc, i mean who writes a NN training code from scratch in this day and age of AI ??

About originality of engines, could you state a rule that will statisfy you ?

[Madeleine Birchfield]

Don’t evade the point. You are fluttering your hands and referring to some “leela backend” that Allie is using. Not true. What there is instead is some highly optimized linear algebra CUDA code written by one individual and contributed to both projects.

Stop making yourself look ridiculous.

Then what exactly makes the CUDA code, written by ankan for the Leela and Allie engines, any different than the NNUE training and evaluation code, written by nodchip for Stockfish, Minic, Igel, Rubichess, Orion and other CPU alpha beta engines? If there isn't any difference, then TCEC and other members of the community should treat one as they treat the other; either allow the likes of Igel to use the Stockfish data/engine in net training, or disallow the likes of Allie/Stein to use the Leela data/engine in net training.

Perhaps 'Leela net player' is incorrect terminology, perhaps we should be calling both Allie and Leela 'ankan NN players' or something to indicate that ankan was the one to contribute to the CUDA backend, in the same way that NNUE is used for nodchip's code.