search

centre-for-humanities-computing / greevaluation

Evaluation workflows for ancient greek language models
MIT License
2 stars 0 forks source link

benchmark speed #14

Open jankounchained opened 1 year ago

jankounchained commented 1 year ago

spacy benchmarks model inference speed in words per minute.

This could be useful info for model comparison. Stanza is painfully slow, udpipe is very fast & our models are probalby somewhere in the middle.

jankounchained commented 1 year ago

problem: might be hard to benchmark the speed of spark nlp. we would have to build a spacy wrapper for it. or steal spacy's speed benchmarking code & adjust it for spark