wink-embeddings-small-en-50d
Version:
Small English 50-dimensional word-embedding dataset compatible with wink-nlp.
13 lines (12 loc) • 389 B
TypeScript
/**
* Convert GloVe txt format (word followed by 50 floats per line) into
* a JSON object mapping word -> number[50]. Optionally limit vocabulary size.
*
* Usage:
* npm run convert:glove -- <path-to-glove.txt> [output.json] [vocabSize]
*
* Example converting 6B.50d:
* npm run convert:glove -- ./glove.6B.50d.txt src/embeddings.json 10000
*/
export {};