spaGO: a work-in-progress yet ready-to-use library for ML/NLP
summary
A production-capable ML/NLP library written entirely in Go that trains and runs models on CPU, with pure Go implementation of MLP, LSTM, BiLSTM+CRF, self-attention/BERT and utilities. It imports PyTorch/HuggingFace weights to fine-tune or perform inference and serves via HTTP JSON or gRPC; demos show question answering and NER returning answers in seconds. Uses gonum assembly, Badger for embeddings and companion projects go-piccolo and go-tokenizers. Current limits: no GPU/TPU, ~60% test coverage, 600+ stars.
AI-written summary · about
chapters (8)
- 0:00Start
- 12:40Building Blocks (continued)
- 16:01Toy Example (c = a+b)
- 17:55Affine transformation
- 20:14Build a Multi-Layer Perceptron (MLP)
- 21:12Put the pieces together
- 23:14NLP: Sequence Labeler (Flair Architecture)
- 24:47NLP: BERT Layer
related talks
28m(Neural) Natural Language Processing in Go
29mThe Go-to Language for AI: Opportunities and Challenges
31mMachine Learning for Gophers
50mDeep Learning with Go
04mGo and data, can they ever be friends?
28mBuilding AI Applications in Go
26mDeep learning in Go - Present and future
29mThe Gopher Orchestra: Distributed Machine Learning with Go