OpenNMT

Health v2 · maintenance only

Adoption not part of the score

7012 stars · 2497 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

OpenNMT is an open-source ecosystem for neural machine translation and sequence learning, with PyTorch (OpenNMT-py) and TensorFlow (OpenNMT-tf) implementations. It supports training and serving Transformer and other seq2seq models, plus LLM finetuning with quantization and LoRA, and is complemented by CTranslate2 for fast inference.

Use cases

• train a custom neural machine translation model
• translate text between languages with a transformer model
• finetune a 7B LLM on a single GPU with 4-bit quantization and LoRA
• build a summarization model from my own data
• serve a translation model behind an API
• train a language model on a custom corpus
• run efficient transformer inference on CPU with CTranslate2
• do seq2seq research with configurable architectures

When to choose

• you need a production-proven, highly configurable NMT training framework
• you want both PyTorch and TensorFlow options for sequence-to-sequence tasks
• you need multi-GPU/multi-node training, beam search, ensembling, and model serving
• you want to finetune or run large language models with quantization

When to avoid

• you want actively developed successors - OpenNMT-py is no longer actively supported in favor of Eole
• you need general-purpose LLM tooling beyond translation/seq2seq where Hugging Face Transformers is a better fit
• you only need inference without training - CTranslate2 alone may suffice

Facets

framework · maturity maintenance

machine-learning deep-learning llm-training llm-inference nlp machine-learning deep-learning large-language-models python neural-machine-translation pytorch tensorflow seq2seq transformer lora quantization ctranslate2 summarization language-modeling natural-language-processing linux macos gpu docker

Member repositories

For agents

markdown · JSON · MCP: product_card(name="OpenNMT/OpenNMT-py")