diff options
Diffstat (limited to 'klm/lm/build_binary.cc')
| -rw-r--r-- | klm/lm/build_binary.cc | 10 |
1 files changed, 5 insertions, 5 deletions
diff --git a/klm/lm/build_binary.cc b/klm/lm/build_binary.cc index ec034640..b340797b 100644 --- a/klm/lm/build_binary.cc +++ b/klm/lm/build_binary.cc @@ -22,9 +22,9 @@ void Usage(const char *name) { "on-disk sort to save memory.\n" "-t is the temporary directory prefix. Default is the output file name.\n" "-m is the amount of memory to use, in MB. Default is 1024MB (1GB).\n\n" -"sorted is like probing but uses a sorted uniform map instead of a hash table.\n" +/*"sorted is like probing but uses a sorted uniform map instead of a hash table.\n" "It uses more memory than trie and is also slower, so there's no real reason to\n" -"use it.\n\n" +"use it.\n\n"*/ "See http://kheafield.com/code/kenlm/benchmark/ for data structure benchmarks.\n" "Passing only an input file will print memory usage of each data structure.\n" "If the ARPA file does not have <unk>, -u sets <unk>'s probability; default 0.0.\n"; @@ -52,13 +52,13 @@ void ShowSizes(const char *file, const lm::ngram::Config &config) { std::size_t probing_size = ProbingModel::Size(counts, config); // probing is always largest so use it to determine number of columns. long int length = std::max<long int>(5, lrint(ceil(log10(probing_size)))); - std::cout << "Memory usage:\ntype "; + std::cout << "Memory estimate:\ntype "; // right align bytes. for (long int i = 0; i < length - 5; ++i) std::cout << ' '; std::cout << "bytes\n" "probing " << std::setw(length) << probing_size << " assuming -p " << config.probing_multiplier << "\n" - "trie " << std::setw(length) << TrieModel::Size(counts, config) << "\n" - "sorted " << std::setw(length) << SortedModel::Size(counts, config) << "\n"; + "trie " << std::setw(length) << TrieModel::Size(counts, config) << "\n"; +/* "sorted " << std::setw(length) << SortedModel::Size(counts, config) << "\n";*/ } } // namespace ngram |