Overcoming Vocabulary Constraints with Pixel-level Fallback
AuthorsJonas F. Lotz†**, Hendra Setiawan, Stephan Peitz, Yova Kementchedjhieva‡
AuthorsJonas F. Lotz†**, Hendra Setiawan, Stephan Peitz, Yova Kementchedjhieva‡
Subword tokenization requires balancing computational efficiency and vocabulary coverage, which often leads to suboptimal performance on languages and scripts not prioritized during training. We propose to augment pretrained language models with a vocabulary-free encoder that generates input embeddings from text rendered as pixels. Through experiments on English-centric language models, we demonstrate that our approach substantially improves machine translation performance and facilitates effective cross-lingual transfer, outperforming tokenizer-based methods. Furthermore, we find that pixel-based representations outperform byte-level approaches and standard vocabulary expansion. Our approach enhances the multilingual capabilities of monolingual language models without extensive retraining and reduces decoding latency via input compression.
February 7, 2025research area Methods and Algorithmsconference ICLR
As language models grow ever larger, so do their vocabularies. This has shifted the memory footprint of LLMs during training disproportionately to one single layer: the cross-entropy in the loss computation. Cross-entropy builds up a logit matrix with entries for each pair of input tokens and vocabulary items and, for small models, consumes an order of magnitude more memory than the rest of the LLM combined. We propose Cut Cross-Entropy (CCE), a...
May 23, 2022research area Privacy, research area Speech and Natural Language Processingconference ACL
Federated learning with differential privacy, i.e. private federated learning (PFL), makes it possible to train models on private data distributed across users’ devices without harming privacy. PFL is efficient for models, such as neural networks, that have a fixed number of parameters, and thus a fixed-dimensional gradient vector. Such models include neural-net language models, but not tokenizers, the topic of this work. Training a tokenizer...