Exponentially Faster Language Modelling
This addresses the problem of slow inference speeds in language models for AI practitioners, though it is incremental as it builds on existing BERT architectures.
The paper tackles the inefficiency of language models by introducing UltraFastBERT, a variant that uses only 0.3% of neurons during inference while matching the performance of similar BERT models, achieving up to 78x speedup on CPU and 40x on PyTorch.
Language models only really need to use an exponential fraction of their neurons for individual inferences. As proof, we present UltraFastBERT, a BERT variant that uses 0.3% of its neurons during inference while performing on par with similar BERT models. UltraFastBERT selectively engages just 12 out of 4095 neurons for each layer inference. This is achieved by replacing feedforward networks with fast feedforward networks (FFFs). While no truly efficient implementation currently exists to unlock the full acceleration potential of conditional neural execution, we provide high-level CPU code achieving 78x speedup over the optimized baseline feedforward implementation, and a PyTorch implementation delivering 40x speedup over the equivalent batched feedforward inference. We publish our training code, benchmarking setup, and model weights.