Yes. There's big arrays, one for each order. These are concatenated together, one after each other. The normal (and current) storage order is 1 2 3 4 5 (for a 5-gram model). Revisions [4037,4040) concatenate arrays as 1 4 3 2 5 because I reversed the iteration order of the loop that assigns storage space for mid-order n-grams without enough thought. So each revision is internally consistent, but using binary files across revisions will (hopefully) segfault.
Sorry about the 1-day confusion, Kenneth On 06/27/11 21:05, Tom Hoar wrote: > Ken, > > To clarify, rev 4041 returns to the file format compatible with rev's > 4036 and before? > > Tom > > > > On Mon, 27 Jun 2011 17:29:58 -0400, Kenneth Heafield > <[email protected]> wrote: >> Folks, >> >> Don't use revisions [4037,4040) with the trie model. I accidentally >> changed the file format and you'll get segfaults on existing binary >> files. Also, the binary files it builds are corrupt. This doesn't sole >> Tom Hoar's problem because his segfault came before revision 4037. >> >> Kenneth >> _______________________________________________ >> Moses-support mailing list >> [email protected] >> http://mailman.mit.edu/mailman/listinfo/moses-support > _______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
