|
|
|
|
|
|
by anamax
5581 days ago
|
|
|
> The algorithm they outline is exponential. They claim that it's actually roughly linear (given some fairly obvious optimizations) and provide a table of measurements that support said claim. http://arxiv.org/PS_cache/arxiv/pdf/1010/1010.5023v1.pdf Number of tokens Parsing time 4,944 10 ms 42,346 43 ms 390,553 326 ms 5,049,213 3.9 s 22,155,402 17.1 s |
|
|