Bit is all we need: binary normalized neural networks

by PaulHoule • 10 months ago

101 54 10 months ago

modeless• 10 months ago

each parameter exists in two forms simultaneously during training: a full-precision 32-bit floating-point value (p) used for gradient updates, and its binarized counterpart (pb) used for forward computations

So this is only for inference. Also activations aren't quantized, I think?

nighthawk454• 10 months ago

Yeah, but it’s ’quantization aware’ during training too, which presumably is what allows the quantization at inference to work

benob• 10 months ago

I wonder if one could store only the binary representation at training and sample a floating point representation (both weights and gradient) during backprop.

adastra22• 10 months ago

Back propagation on random data that is then thrown away would be pretty useless.

jampekka• 10 months ago

Also activations aren't quantized, I think?

The very last conclusion: "Future work will focus on the implementation of binary normalization layers using single-bit arrays operations, as well as on quantizing layer activations to 8 or 16-bit precision. These improvements are expected to further enhance the efficiency and performance of the binary neural network models."

kouteiheika• 10 months ago

You don't necessarily have to store the parameters in fp32 for gradient updates; I experimented with it and got it working (all parameter full fine-tuning) with parameters being as low as 3-bit (a little bit more than 3-bit, because the block-wise scales were higher precision), which is essentially as low as you can go before "normal" training starts breaking down.

noosphr• 10 months ago

Yes, that's been the downside of these forever.

If you use quantized differentiation you can get away with using integers for gradient updates. Explaining how takes a paper and in the end it doesn't even work very well.

At university, way back at the end of the last AI winter, I ended up using genetic algorithms to train the models. It was very interesting because weights were trained along with hyper parameters. It was no where near practical because gradient descent is so much better at getting real world results in reasonable time frames - surprisingly because it's more memory efficient.

AmazingTurtle• 10 months ago

I'm gonna refer to this one here: https://news.ycombinator.com/item?id=45361007

s20n• 10 months ago

Attention Is All You Need - The Beatles ft. Charlie Puth

JKCalhoun• 10 months ago

b/w "All You Need is Love".

amelius2• 10 months ago

Attention is all Google needs. Apparently.

I'm sick of BigTech fighting for my attention.

mnky9800n• 10 months ago

Yes calling your paper this now makes me think your paper has no interesting results. It is kind of the opposite of what’s intended.

gloomyday• 10 months ago

This naming trend has been going for 8 years. Incredible.

IshKebab• 10 months ago

It's on my naughty list, together with "... considered harmful", "The unreasonable effectiveness of ...", "... for fun and profit", "Falsehoods programmers believe about ...", "The rise and fall of ...".

fxtentacle2• 10 months ago

These techniques are not new. And the reason why they’re usually not used is on page 9 in the paper. They require about 10x as many training iterations.

typpilol• 10 months ago

Yea I saw that training perplexity and thought hmmm...

shomp2• 10 months ago

Turns out using floats is a feature and not a bug?

Dylan168074• 10 months ago

No, I don't think so, in that I don't think anyone has ever called that a bug.

personalityson• 10 months ago

Unless each iteration is 90% faster

amelius2• 10 months ago

This.

In fact, it can be slower because hardware is probably not optimized for the 1-bit case, so there may be a lot of low-hanging fruit for hardware designers and we may see improvements in the next iteration of hardware.

fxtentacle2• 10 months ago

In this paper, each iteration has to be slower. Because they need to calculate both their new method (which may be faster) and also the traditional method (because they need a float gradient). And old+new will always be slower than just old.

PaulHoule2• 10 months ago

When I was working for startups trying to develop foundation models circa 2015 we were concerned with training more than inference.

Today with models that are actually useful training costs matters much less than inference costs. A 10x increase in training costs is not necessarily prohibitive if you get a 10x decrease in inference costs.

nickpsecurity5• 10 months ago

I still don't have a GPT3-class model that was trained without copyright infringement. I'd have so many uses for it from research to production. What's stopping me is the $30 million training cost for 180B models. Even a 30B like Mosaic cost over a million dollars.

So, I strongly disagree unless we're talking about the five or six companies that already spend tens of millions on training and keep repeating that. Outside of them, the medium to large models are done infrequently or one off by a small number of other companies. Then, most of us are stuck with their pretraining efforts because we can't afford it ourselves.

On my end, I'd rather see a model that drops pretraining costs to almost nothing but costs 10-32x more to do inference. My uses would produce mere MB of output vs hundreds of GB to TB that pretraining requires. A competitive use that costs 32x current prices would probably be profitable for me. Optimizations, which are plentiful for inference, might bring it down further.

pixelpoet3• 10 months ago

The critical "1" is missing from the title...

Dylan168074• 10 months ago

"Bit" being singular gets the intent across just fine.

pixelpoet3• 10 months ago

Disagree

user823749• 10 months ago

Yes, just like in "16 bit integer". No confusion at all.

jongjong2• 10 months ago

This reminds me if my university days. For one of the assignments, we had to write our own ANN from scratch for handwriting recognition and we implemented a step activation function because that was easier than sigmoid; basically each layer would output one or zero though I guess the weights themselves were scalars. It's just the node outputs which were 1 or 0... But this was convenient because the output of the final layer could be interpreted as a binary which could be converted straight into an ASCII character for comparison and backpropagation.

meindnoch• 10 months ago

could be interpreted as a binary which could be converted straight into an ASCII character for comparison and backpropagation.

There's nothing to backpropagate with a step function. The derivative is zero everywhere.

steppi2• 10 months ago

It sounds like jongjong was probably using surrogate gradients. You keep the step activation in the forward pass but replace with a smooth approximation in the backwards pass.

forntentacl• 10 months ago

This paper ignores 50+ years of research in the domain of quantized networks, quantized training algorithms, and reaches wrong conclusions out of sheer ignorance.

TLDR abstract of a draft paper I wrote years ago, for those interested in the real limits of quantized networks:

We investigate the storage capacity of single‐layer threshold neurons under three synaptic precision regimes—binary (1‐bit), ternary (≈1.585‐bit), and quaternary (2‐bit)—from both information‐theoretic and algorithmic standpoints. While the Gardner bound stipulates maximal loads of α=0.83, 1.5 and 2.0 patterns per weight for the three regimes, practical algorithms only reach α_alg≈0.72, 1.0 and 2.0, respectively. By converting these densities into storage‐efficiency metrics—bits of synaptic memory per stored pattern—we demonstrate that only quaternary weights achieve the theoretical optimum in realistic settings, requiring exactly 1 bit of memory per pattern. Binary and ternary schemes incur 39 % and 58 % overheads, respectively.

naasking• 10 months ago

Is this actually equivalent to classical forms of quantization though? The paper has extensive discussion of quantization on page 2 and 3. This paper is not just a rehash of earlier work, but pushes the single bit precision to more parts of the system.

thijson• 10 months ago

How does this compare to:

https://arxiv.org/pdf/1811.11431

Bit is all we need: binary normalized neural networks

Related Stories