The recently proposed Forgetting Transformer (FoX) incorporates a forget gate into softmax attention and has shown consistently better or on-par performance compared to the standard RoPE-based Transformer. Notably, many attention heads in FoX tend to forget quickly, causing their output at each timestep to rely primarily on local context. Based on this observation, we propose Adaptive Computation Pruning (ACP) for FoX, a method that dynamically prunes computations involving input-output dependencies that are strongly decayed by the forget gate. In particular, our method performs provably safe pruning via a dynamically set pruning threshold that guarantees the pruned attention weights are negligible. We apply ACP to language model pretraining with FoX and show it consistently reduces the number of FLOPs and memory accesses in softmax attention by around 70% across different model sizes and context lengths, resulting in a roughly 50% to 70% reduction in attention runtime (or a 2-3$\times$ speedup) and a roughly 10% to 40% increase in end-to-end training throughput. Furthermore, longer context lengths yield greater computational savings. All these speed improvements are achieved without any performance degradation. Our code is available at https://github.com/zhixuan-lin/forgetting-transformer.

An essential component of modern recurrent sequence models is the forget gate. While Transformers do not have an explicit recurrent form, we show that a forget gate can be naturally incorporated into Transformers by down-weighting the unnormalized attention scores in a data-dependent way. We name this attention mechanism the Forgetting Attention and the resulting model the Forgetting Transformer (FoX). We show that FoX outperforms the Transformer on long-context language modeling, length extrapolation, and short-context downstream tasks, while performing on par with the Transformer on long-context downstream tasks. Moreover, it is compatible with the FlashAttention algorithm and does not require any positional embeddings. Several analyses, including the needle-in-the-haystack test, show that FoX also retains the Transformer's superior long-context capabilities over recurrent sequence models such as Mamba-2, HGRN2, and DeltaNet. We also introduce a "Pro" block design that incorporates some common architectural components in recurrent sequence models and find it significantly improves the performance of both FoX and the Transformer. Our code is available at https://github.com/zhixuan-lin/forgetting-transformer.

Transformer-based open-domain dialog models have become increasingly popular in recent years. These models typically represent context as a concatenation of a dialog history. However, there is no criterion to decide how many utterances should be kept adequate in a context. We try to figure out how the choice of context length affects the model. We experiment on three questions from coarse to fine: (i) Does longer context help model training? (ii) Is it necessary to change the training context length when dealing with dialogs of different context lengths? (iii) Do different dialog samples have the same preference for context length? Our experimental results show that context length, an often overlooked setting, deserves attention when implementing Transformer-based dialog models.

Accommodating long sequences efficiently in autoregressive Transformers, especially within an extended context window, poses significant challenges due to the quadratic computational complexity and substantial KV memory requirements inherent in self-attention mechanisms. In this work, we introduce SPARSEK Attention, a novel sparse attention mechanism designed to overcome these computational and memory obstacles while maintaining performance. Our approach integrates a scoring network and a differentiable top-k mask operator, SPARSEK, to select a constant number of KV pairs for each query, thereby enabling gradient-based optimization. As a result, SPARSEK Attention offers linear time complexity and constant memory footprint during generation. Experimental results reveal that SPARSEK Attention outperforms previous sparse attention methods and provides significant speed improvements during both training and inference, particularly in language modeling and downstream tasks. Furthermore, our method can be seamlessly integrated into pre-trained Large Language Models (LLMs) with minimal fine-tuning, offering a practical solution for effectively managing long-range dependencies in diverse applications.