5 ms·

The way I understood it, the author is saying that, with this change, big values disappear, and we can then use fewer bits to encode the output of transformers,

The way I understood it, the author is saying that, with this change, big values disappear, and we can then use fewer bits to encode the output of transformers, which means reducing the memory requirements of the network. Memory being the limiting factor to running models large, this would be a big deal.

AlanSE 3y ago
Yeah, good to bring it back to the original point. Reading the article felt exciting, but in hindsight I am now missing a key detail. The equations all seem to be matrix operations with a fixed number of rows / columns (you can take me as a real layman here). Unless you change that, I don't understand _how_ you can reduce memory needs. Granted, I'm probably putting my foot in my mouth not understanding transformers.
- sudosysgen 3y ago
  You can reduce the number of bits per float (scalar).
- jablongo 3y ago
  It has to do with the precision of the values stored in those rows and columns. If they could be coerced into a narrower range (without losing information) then we could effectively store them each with 8 bits or something. The +1 prevents blowups when the denominator in its current form approaches 0, and without those blowups, then we can use less bits, in theory.
  - cycomanic 3y ago
    That is only true if the using the new softmax changes the dynamic range of the values. We are using floating point not fixed point. So if before our values went from 1 to 5000 and now they go from 0.0002 to 1 we still have the same dynamic range and so still need the same resolution.
    - IanCal 3y ago
      The quantized versions are not floats but ints.
- zamalek 3y ago
  More ELI5 than the other comments. Considering the softmax network: During quantization we find that values in the network vary from 0->5000, but 95% of values are <100. Quantizing this to 8bits would mean that our values would be in increments of about 20. Remembering that 95% of our values are below 100, we would only have about 5 discrete values for 95% of our values - so we would be losing a lot of "resolution" (entropy/information). For example (assuming rounding is used), an original value of 19 would be quantized to 20 and 30 would be quantized to 40. The original values differ by 11, but the quantized values differ by 20! This is where exotic encodings come into play. We might try to use a logarithmic scheme, for example. This would result in higher value densities at lower values - but we would probably still waste bits and it would require more APU cycles. Now switch to the softmax1 network: The range of values is less important than the distribution - instead of 95% of the values falling in a small range, we would see the values more evenly spread out. Assuming that the range is now 105 (so the 5% outlying neurons from the softmax network are still >100), we would have 243 values to represent everything under 100. The same example with 19 and 30 would result in 19.27 and 30.34 respectively, a difference of 11.07 - which is very close to the unquantized difference of 11. We have retained more information in the quantized version of the network. Information is lost either way, but what's important is how much information is lost. The reason that the large values appear is because the heads attempt to "scream really loud" when they are certain that they are right. This is an emergent behavior due to softmax - it ironically sucks at paying attention to a few of the heads: it boosts the volume of the heads that are trying to abstain, and mutes the volume of the heads that are trying to vote.
  - samwillis 3y ago
    If I'm following correctly, does this mean that with this change along with a model being quantized, we could see models that are 5% the size (on file system) and memory usage but almost identical in output?
    - zamalek 3y ago
      The vales are selected were arbitrary. The size reduction will be 32bits/8bits - so it will be 4 times smaller.
- uoaei 3y ago
  The activations (outputs) of one layer must be encoded in the same way as the weights of that layer as well as the weights of the next layer or the computation fails (unless you manage to write clever kernels for doing math at different levels of precision simultaneously, but even then you're introducing even more lossiness than just using a binary representation for those values). Example: multiplying a bunch of float16s together gives you a float16. That is passed on to the next layer of float16s. Why should forcing the output of the first step to be float8 confer any advantage here? The only way I can see this argument working is if you make all the layers float8 too, and the reason you can do that is that the output of the first step can be faithfully represented as float8 because it doesn't ever blow up. If that's what the author is saying, it wasn't very clear.
bambax 3y ago
> The Qualcomm AI researchers found that 97%+ of outlier activations in LLMs occur in whitespace and punctuation positions. This is striking. If true, why not try to ignore whitespace and puctuation? In old Latin, scripto continua [1] was a way to write continuously, for the exact same reason: to save space. Other modern languages still do that, and are no less parseable. Granted, it's unlikely a commercial LLM would become popular if it produced output without spaces or punctuation; but an open source one that promised to be much more compressible, and therefore work on smaller machines, might be super useful. It's not hard for a human to add spaces afterwards. It used to be a job for beginning journalists at the time of telex machines: press releases were sent in all caps without spaces, and interns were tasked with adding slashes between words. In French it was called "bâtonner les dépêches" (literally: add sticks to press releases -- not sure about the idiomatic English translation). [1] https://simple.wikipedia.org/wiki/Scriptio_continua https://simple.wikipedia.org/wiki/Scriptio_continua
- colordrops 3y ago
  Seems you could make a pipeline where a much simpler model adds spaces and punctuation to output from the main model.
  - samwillis 3y ago
    I suspect punctuation adds significant meaning to the models, that could be why so much computation is applied to it. That's not to say a pipeline couldn't be effective.
  - bambax 3y ago
    Yes I was thinking about that, it should be quite easy afterwards.
  - GTP 3y ago
    For the spaces and for some (maybe most?) languages you don't even need a NN to add spaces: as words made of two or more words aren't that common, and when those occur you probably want to use the composite one, it boils down to start from the beginning of the text and look in a dictionary what's the longest string that is a valid word. The only language that I know of that uses a lot of composite words (I mean words made by sticking two or more words togheter) is German, but I think that looking for the longest sequence occurring in a dictionary would be correct most of the times.

AlanSE 3y ago

Yeah, good to bring it back to the original point. Reading the article felt exciting, but in hindsight I am now missing a key detail. The equations all seem to be matrix operations with a fixed number of rows / columns (you can take me as a real layman here). Unless you change that, I don't understand _how_ you can reduce memory needs. Granted, I'm probably putting my foot in my mouth not understanding transformers.

sudosysgen 3y ago
You can reduce the number of bits per float (scalar).
jablongo 3y ago
It has to do with the precision of the values stored in those rows and columns. If they could be coerced into a narrower range (without losing information) then we could effectively store them each with 8 bits or something. The +1 prevents blowups when the denominator in its current form approaches 0, and without those blowups, then we can use less bits, in theory.
- cycomanic 3y ago
  That is only true if the using the new softmax changes the dynamic range of the values. We are using floating point not fixed point. So if before our values went from 1 to 5000 and now they go from 0.0002 to 1 we still have the same dynamic range and so still need the same resolution.
  - IanCal 3y ago
    The quantized versions are not floats but ints.
zamalek 3y ago
More ELI5 than the other comments. Considering the softmax network: During quantization we find that values in the network vary from 0->5000, but 95% of values are <100. Quantizing this to 8bits would mean that our values would be in increments of about 20. Remembering that 95% of our values are below 100, we would only have about 5 discrete values for 95% of our values - so we would be losing a lot of "resolution" (entropy/information). For example (assuming rounding is used), an original value of 19 would be quantized to 20 and 30 would be quantized to 40. The original values differ by 11, but the quantized values differ by 20! This is where exotic encodings come into play. We might try to use a logarithmic scheme, for example. This would result in higher value densities at lower values - but we would probably still waste bits and it would require more APU cycles. Now switch to the softmax1 network: The range of values is less important than the distribution - instead of 95% of the values falling in a small range, we would see the values more evenly spread out. Assuming that the range is now 105 (so the 5% outlying neurons from the softmax network are still >100), we would have 243 values to represent everything under 100. The same example with 19 and 30 would result in 19.27 and 30.34 respectively, a difference of 11.07 - which is very close to the unquantized difference of 11. We have retained more information in the quantized version of the network. Information is lost either way, but what's important is how much information is lost. The reason that the large values appear is because the heads attempt to "scream really loud" when they are certain that they are right. This is an emergent behavior due to softmax - it ironically sucks at paying attention to a few of the heads: it boosts the volume of the heads that are trying to abstain, and mutes the volume of the heads that are trying to vote.
- samwillis 3y ago
  If I'm following correctly, does this mean that with this change along with a model being quantized, we could see models that are 5% the size (on file system) and memory usage but almost identical in output?
  - zamalek 3y ago
    The vales are selected were arbitrary. The size reduction will be 32bits/8bits - so it will be 4 times smaller.
uoaei 3y ago
The activations (outputs) of one layer must be encoded in the same way as the weights of that layer as well as the weights of the next layer or the computation fails (unless you manage to write clever kernels for doing math at different levels of precision simultaneously, but even then you're introducing even more lossiness than just using a binary representation for those values). Example: multiplying a bunch of float16s together gives you a float16. That is passed on to the next layer of float16s. Why should forcing the output of the first step to be float8 confer any advantage here? The only way I can see this argument working is if you make all the layers float8 too, and the reason you can do that is that the output of the first step can be faithfully represented as float8 because it doesn't ever blow up. If that's what the author is saying, it wasn't very clear.

bambax 3y ago

> The Qualcomm AI researchers found that 97%+ of outlier activations in LLMs occur in whitespace and punctuation positions. This is striking. If true, why not try to ignore whitespace and puctuation? In old Latin, scripto continua [1] was a way to write continuously, for the exact same reason: to save space. Other modern languages still do that, and are no less parseable. Granted, it's unlikely a commercial LLM would become popular if it produced output without spaces or punctuation; but an open source one that promised to be much more compressible, and therefore work on smaller machines, might be super useful. It's not hard for a human to add spaces afterwards. It used to be a job for beginning journalists at the time of telex machines: press releases were sent in all caps without spaces, and interns were tasked with adding slashes between words. In French it was called "bâtonner les dépêches" (literally: add sticks to press releases -- not sure about the idiomatic English translation). [1] https://simple.wikipedia.org/wiki/Scriptio_continua https://simple.wikipedia.org/wiki/Scriptio_continua

colordrops 3y ago
Seems you could make a pipeline where a much simpler model adds spaces and punctuation to output from the main model.
- samwillis 3y ago
  I suspect punctuation adds significant meaning to the models, that could be why so much computation is applied to it. That's not to say a pipeline couldn't be effective.
- bambax 3y ago
  Yes I was thinking about that, it should be quite easy afterwards.
- GTP 3y ago
  For the spaces and for some (maybe most?) languages you don't even need a NN to add spaces: as words made of two or more words aren't that common, and when those occur you probably want to use the composite one, it boils down to start from the beginning of the text and look in a dictionary what's the longest string that is a valid word. The only language that I know of that uses a lot of composite words (I mean words made by sticking two or more words togheter) is German, but I think that looking for the longest sequence occurring in a dictionary would be correct most of the times.