diff options
author | trevor.cohn <trevor.cohn@ec762483-ff6d-05da-a07a-a48fb63a330f> | 2010-07-21 18:08:59 +0000 |
---|---|---|
committer | trevor.cohn <trevor.cohn@ec762483-ff6d-05da-a07a-a48fb63a330f> | 2010-07-21 18:08:59 +0000 |
commit | 6b6679b96cb19883319f05c36f0057f4b0188bd6 (patch) | |
tree | 652e60d88c7165a153442e0a50e89f1e956a2d1c /gi/pipeline | |
parent | a592b3a5ed2dc96c62d9fe7408948d66b97e1aec (diff) |
Agreement model flag
git-svn-id: https://ws10smt.googlecode.com/svn/trunk@358 ec762483-ff6d-05da-a07a-a48fb63a330f
Diffstat (limited to 'gi/pipeline')
-rwxr-xr-x | gi/pipeline/local-gi-pipeline.pl | 26 |
1 files changed, 17 insertions, 9 deletions
diff --git a/gi/pipeline/local-gi-pipeline.pl b/gi/pipeline/local-gi-pipeline.pl index 6b58ab7c..9cffc492 100755 --- a/gi/pipeline/local-gi-pipeline.pl +++ b/gi/pipeline/local-gi-pipeline.pl @@ -48,7 +48,9 @@ my $PATCH_CORPUS = "$SCRIPT_DIR/scripts/patch-corpus.pl"; my $EXTRACTOR = "$EXTOOLS/extractor"; my $TOPIC_TRAIN = "$PYPTOOLS/pyp-contexts-train"; -assert_exec($PATCH_CORPUS, $SORT_KEYS, $REDUCER, $EXTRACTOR, $PYP_TOPICS_TRAIN, $S2L, $C2D, $TOPIC_TRAIN); +assert_exec($PATCH_CORPUS, $SORT_KEYS, $REDUCER, $EXTRACTOR, #$PYP_TOPICS_TRAIN, + $S2L, $C2D #, $TOPIC_TRAIN + ); my $BACKOFF_GRAMMAR; my $DEFAULT_CAT; @@ -142,7 +144,7 @@ if (lc($MODEL) eq "pyp") { } else { topic_train(); } -} elsif (lc($MODEL) eq "prem") { +} elsif (lc($MODEL) =~ /pr|em|agree/) prem_train(); } else { die "Unsupported model type: $MODEL. Must be one of PYP or PREM.\n"; } if($HIER_CAT) { @@ -191,12 +193,12 @@ sub context_dir { sub cluster_dir { if (lc($MODEL) eq "pyp") { return context_dir() . ".PYP.t$NUM_TOPICS.s$NUM_SAMPLES"; - } elsif (lc($MODEL) eq "prem") { - if ($PR_SCALE_P == 0 && $PR_SCALE_C == 0) { - return context_dir() . ".EM.t$NUM_TOPICS.i$NUM_ITERS"; - } else { - return context_dir() . ".PR.t$NUM_TOPICS.i$NUM_ITERS.sp$PR_SCALE_P.sc$PR_SCALE_C"; - } + } elsif (lc($MODEL) eq "em") { + return context_dir() . ".EM.t$NUM_TOPICS.i$NUM_ITERS"; + } elsif (lc($MODEL) eq "pr") { + return context_dir() . ".PR.t$NUM_TOPICS.i$NUM_ITERS.sp$PR_SCALE_P.sc$PR_SCALE_C"; + } elsif (lc($MODEL) eq "agree") { + return context_dir() . ".AGREE.t$NUM_TOPICS.i$NUM_ITERS"; } } @@ -277,7 +279,13 @@ sub prem_train { if (-e $OUT_CLUSTERS) { print STDERR "$OUT_CLUSTERS exists, reusing...\n"; } else { - safesystem("$PREM_TRAIN --in $IN_CONTEXTS --topics $NUM_TOPICS --out $OUT_CLUSTERS --iterations $NUM_ITERS --scale-phrase $PR_SCALE_P --scale-context $PR_SCALE_C $PR_FLAGS") or die "Topic training failed.\n"; + my $opts = ""; + if (lc($MODEL) eq "pr") { + $opts = "--scale-phrase $PR_SCALE_P --scale-context $PR_SCALE_C"; + } elsif (lc($MODEL) eq "agree") { + $opts = "--agree-direction"; + } + safesystem("$PREM_TRAIN --in $IN_CONTEXTS --topics $NUM_TOPICS --out $OUT_CLUSTERS --iterations $NUM_ITERS $opts $PR_FLAGS") or die "Topic training failed.\n"; } } |