Scaling Laws of Decoder-Only Models on the Multilingual Machine Translation Task
This work addresses the scaling behavior of decoder-only models for multilingual machine translation, an incremental contribution as it extends known scaling laws to a less-studied model type in this domain.
The study investigated scaling laws for decoder-only models in multilingual machine translation, finding that their loss can be estimated similarly to large language models but struggles with generalization to very large models or different data distributions, with experiments on models from 70M to 7B parameters showing similar test loss improvements from scaling depth or width but differing efficiency impacts.
Recent studies have showcased remarkable capabilities of decoder-only models in many NLP tasks, including translation. Yet, the machine translation field has been largely dominated by encoder-decoder models based on the Transformer architecture. As a consequence, scaling laws of encoder-decoder models for neural machine translation have already been well studied, but decoder-only models have received less attention. This work explores the scaling laws of decoder-only models on the multilingual and multidomain translation task. We trained a collection of six decoder-only models, ranging from 70M to 7B parameters, on a sentence-level, multilingual and multidomain dataset. We conducted a series of experiments showing that the loss of decoder-only models can be estimated using a scaling law similar to the one discovered for large language models, but we also show that this scaling law has difficulties to generalize to too large models or to a different data distribution. We also study different scaling methods and show that scaling the depth and the width of a model lead to similar test loss improvements, but with different impact on the model's efficiency.