Tag

fairseq is an open-source sequence modeling toolkit from Facebook AI Research (FAIR), implemented in Python on top of PyTorch. It provides reference implementations for a wide range of sequence models (Transformer, LSTM, Conv, wav2vec, wav2vec 2.0, etc.) and supports tasks such as machine translation, summarization, language modeling, and speech processing. Key features include multi-GPU and distributed training, fast generation (beam search, sampling, diverse beam), mixed-precision training, parameter/optimizer sharding, and many pre-trained models and examples. The project is MIT-licensed and documented at readthedocs.

Buzz is an open-source desktop and CLI tool that transcribes and translates audio offline using OpenAI's Whisper. It supports macOS, Windows and Linux, offers GUI and command-line interfaces, can be installed via native installers, Flatpak/Snap, winget or PyPI, and supports GPU acceleration via PyTorch/CUDA.

VideoCaptioner is an AI-powered video subtitling assistant that combines ASR (local or cloud) with LLM-based subtitle segmentation, correction and translation. It supports offline GPU transcription, concurrent chunk transcription, VAD, speaker-aware processing, batch subtitling and one-click subtitle-to-video synthesis, with both GUI and CLI options.

BabelDOC is a PDF-focused document translation and bilingual comparison library/CLI that converts scientific papers and other PDFs (primarily English→Chinese) using OpenAI-compatible LLMs. It provides CLI and Python APIs, supports integration with self-hosted pipelines (e.g., PDFMathTranslate-next), offers online service through Immersive Translate, and includes many PDF-processing and translation options (OCR, glossary extraction, bilingual output, watermark controls).

This paper introduces an attention-based encoder–decoder NMT architecture that learns soft alignments between source and target words while translating, eliminating the fixed-length bottleneck of earlier seq2seq models. The approach substantially improves BLEU, especially on long sentences, and matches phrase-based SMT on English-French without additional hand-engineered features. The attention mechanism it proposes became the foundation for virtually all subsequent NMT systems and inspired attention-centric models like the Transformer, reshaping machine translation and sequence modeling across NLP.