diff options
Diffstat (limited to 'train_test_split')
-rwxr-xr-x | train_test_split | 50 |
1 files changed, 50 insertions, 0 deletions
diff --git a/train_test_split b/train_test_split new file mode 100755 index 0000000..db56de9 --- /dev/null +++ b/train_test_split @@ -0,0 +1,50 @@ +#!/usr/bin/env ruby + +require 'zipf' +require 'trollop' + +cfg = Trollop::options do + opt :foreign, "foreign file", :type => :string, :required => true + opt :english, "english file", :type => :string, :required => true + opt :size, "one size", :type => :int, :required => true + opt :repeat, "number of repetitions", :type => :int, :default => 1 + opt :prefix, "prefix for output files", :type => :string +end +fn = cfg[:foreign] +fn_ext = fn.split('.').last +f = ReadFile.readlines fn +en = cfg[:english] +en_ext = en.split('.').last +e = ReadFile.readlines en +size = cfg[:size] +nlines_f = `wc -l #{fn}`.split()[0].to_i +nlines_e = `wc -l #{en}`.split()[0].to_i +if nlines_f != nlines_e + STDERR.write "Unbalanced files (#{nlines_f} vs. #{nlines_e}), exiting!\n" + exit 1 +end + +prefix = cfg[:prefix] +a = (0..nlines_e-1).to_a +i = 0 +cfg[:repeat].times { + b = a.sample(size) + ax = a.reject{|j| b.include? j} + `mkdir split_#{i}` + new_f = WriteFile.new "split_#{i}/#{prefix}.train.#{i}.#{fn_ext}" + new_e = WriteFile.new "split_#{i}/#{prefix}.train.#{i}.#{en_ext}" + ax.each { |j| + new_f.write f[j] + new_e.write e[j] + } + new_f.close; new_e.close + new_f = WriteFile.new "split_#{i}/#{prefix}.test.#{i}.#{fn_ext}" + new_e = WriteFile.new "split_#{i}/#{prefix}.test.#{i}.#{en_ext}" + b.each { |j| + new_f.write f[j] + new_e.write e[j] + } + new_f.close; new_e.close + i += 1 +} + |