Sockeye 3: Fast Neural Machine Translation with PyTorch
This provides a more efficient toolkit for researchers and practitioners in machine translation, enabling faster experimentation and deployment, though it is incremental as an update to an existing toolkit.
The paper tackles the need for faster neural machine translation by introducing Sockeye 3, a toolkit based on PyTorch that achieves up to 126% faster performance on GPUs and up to 292% faster on CPUs compared to other implementations.
Sockeye 3 is the latest version of the Sockeye toolkit for Neural Machine Translation (NMT). Now based on PyTorch, Sockeye 3 provides faster model implementations and more advanced features with a further streamlined codebase. This enables broader experimentation with faster iteration, efficient training of stronger and faster models, and the flexibility to move new ideas quickly from research to production. When running comparable models, Sockeye 3 is up to 126% faster than other PyTorch implementations on GPUs and up to 292% faster on CPUs. Sockeye 3 is open source software released under the Apache 2.0 license.