Comment by cs702

Comment by cs702 3 hours ago

7 replies

View on Hacker News

> Who out there is working on ... infinite history?

Many people are still working on improving RNNs, mostly in academia. Examples off the top of my head:

* RWKV: https://arxiv.org/abs/2006.16236 / https://arxiv.org/abs/2404.05892 https://arxiv.org/abs/2305.13048

* Linear attention: https://arxiv.org/abs/2503.14456

* State space models: https://arxiv.org/abs/2312.00752 / https://arxiv.org/abs/2405.21060

* Linear RNNs: https://arxiv.org/abs/2410.01201

Industry OTOH has gone all-in on Transformers.

jmward01 2 hours ago

RNNs have two huge issues: - long context. Recurrence degrades the signal for the same reason that 'deep' nn architectures don't go much past 3-4 layers before you need residual connections and the like - (this is the big one) training performance is terrible since you can't parallelize them across a sequence like you can with causal masked attn in transformers

On the huge benefit side though you get: - guaranteed state size so perfect batch packing, perfect memory use, easy load/unload from a batch, O(1) of token gen so generally massive performance gains in inference. - unlimited context (well, no need for a concept of a position embedding or similar system)

Taking the best of both worlds is definitely where it is at for the future. An architecture that can train parallelized, has a fixed state size so you can load/unload and patch batches perfectly, unlimited context (with perfect recall), etc etc. That is the real architecture to go for.

Reply View 5 replies
  • zozbot234 2 hours ago

    RNN training cannot be parallelized along the sequence dimension like attention can, but it can still be trained in batches on multiple sequences simultaneously. Given the sizes of modern training sets and the limits on context size for transformer-based models, it's not clear to what extent this is an important limitation nowadays. It may have been more relevant in the early days of attention-based models where being able to do experimental training runs quickly on relatively small sizes of training data may have been important.

    Reply View 2 replies
    • DeveloperErrata 36 minutes ago

      Not quite, most of the recent work on modern RNNs has been addressing this exact limitation. For instance linear attention yields formulations that can be equivalently interpreted either as a parallel operation or a recursive one. The consequence is that these parallelizable versions of RNNs are often "less expressive per param" than their old-school non-parallelizable RNN counterparts, though you could argue that they make up for that in practice by being more powerful per unit of training compute via much better training efficiency.

      Reply View 0 replies
    • jmward01 an hour ago

      To get a similar token/sec in training though you would need to swap batch size and seq length so you could have the massive batch size but then won't you start hitting memory issues with any reasonable sequence length? You would have to create do something similar to a minibatch along the sequence and cut the gradients after a short number of tokens on each sequence. So how will they learn truly long sequences for recall? Or is there a different trick I am missing here?

      Reply View 0 replies
  • cs702 an hour ago

    Linear RNNs overcome both issues. All the RNNs I mentioned are linear RNNs.

    Reply View 1 reply
viraptor 2 hours ago

> Industry OTOH has gone all-in on Transformers.

It's so annoying. Transformers keep improving and recurrent networks are harder to train so until we hit some real wall, companies don't seem eager to diverge. It's like lithium batteries improving easy faster than it was profitable to work on sodium ones, even though we unfortunately want the sodium ones to be better.

Reply View 0 replies