diff options
author | Patrick Simianer <p@simianer.de> | 2011-09-09 15:33:35 +0200 |
---|---|---|
committer | Patrick Simianer <p@simianer.de> | 2011-09-23 19:13:58 +0200 |
commit | edb0cc0cbae1e75e4aeedb6360eab325effe6573 (patch) | |
tree | a2fed4614b88f177f91e88fef3b269fa75e80188 /klm/lm/vocab.hh | |
parent | 2e6ef7cbec77b22ce3d64416a5ada3a6c081f9e2 (diff) |
partial merge, ruleid feature
Diffstat (limited to 'klm/lm/vocab.hh')
-rw-r--r-- | klm/lm/vocab.hh | 4 |
1 files changed, 4 insertions, 0 deletions
diff --git a/klm/lm/vocab.hh b/klm/lm/vocab.hh index c92518e4..9d218fff 100644 --- a/klm/lm/vocab.hh +++ b/klm/lm/vocab.hh @@ -61,6 +61,7 @@ class SortedVocabulary : public base::Vocabulary { } } + // Size for purposes of file writing static size_t Size(std::size_t entries, const Config &config); // Vocab words are [0, Bound()) Only valid after FinishedLoading/LoadedBinary. @@ -77,6 +78,9 @@ class SortedVocabulary : public base::Vocabulary { // Reorders reorder_vocab so that the IDs are sorted. void FinishedLoading(ProbBackoff *reorder_vocab); + // Trie stores the correct counts including <unk> in the header. If this was previously sized based on a count exluding <unk>, padding with 8 bytes will make it the correct size based on a count including <unk>. + std::size_t UnkCountChangePadding() const { return SawUnk() ? 0 : sizeof(uint64_t); } + bool SawUnk() const { return saw_unk_; } void LoadedBinary(int fd, EnumerateVocab *to); |