| ▲ | mbgerring 20 hours ago | |
Also I’m going to go ahead and say that “it’s slightly better than classical NLP for grammar check but requires 10,000x as much compute resources” is not an improvement | ||
| ▲ | bee_rider 9 hours ago | parent [-] | |
The models are already trained… is inference so costly? It’s just a dot product or a matvec or something, right? | ||