merge older changes, more polishing

1 files changed, 11 insertions, 11 deletions

diff --git a/dtrain/test/example/dtrain.ini b/dtrain/test/example/dtrain.ini
index 68173e11..66be6bf2 100644
--- a/dtrain/test/example/dtrain.ini
+++ b/dtrain/test/example/dtrain.ini
@@ -1,20 +1,20 @@
 input=test/example/nc-wmt11.1k.gz    # use '-' for stdin
-output=-                             # a weights file or stdout
-decoder_config=test/example/cdec.ini # ini for cdec
-# these will be printed on each iteration
+output=weights.gz                    # a weights file (add .gz for gzip compression) or STDOUT '-'
+decoder_config=test/example/cdec.ini # config for cdec
+# weights for these features will be printed on each iteration
 print_weights=Glue WordPenalty LanguageModel LanguageModel_OOV PhraseModel_0 PhraseModel_1 PhraseModel_2 PhraseModel_3 PhraseModel_4 PhraseModel_5 PhraseModel_6 PassThrough
 tmp=/tmp
-stop_after=10 # stop iteration after 10 inputs 
+stop_after=100 # stop epoch after 10 inputs
 
 # interesting stuff
-epochs=3                # run over input 3 times
-k=200                   # use 100best lists
-N=4                     # optimize (approx) BLEU4
+epochs=100                # run over input 3 times
+k=100                   # use 100best lists
+N=4                    # optimize (approx) BLEU4
 learning_rate=0.0001    # learning rate
-gamma=0.00001           # use SVM reg
-scorer=stupid_bleu      # use stupid BLEU+1 approx.
+gamma=0           # use SVM reg
+scorer=smooth_bleu      # use smooth BLEU of (Liang et al. '06)
 sample_from=kbest       # use kbest lists (as opposed to forest)
-filter=uniq             # only uniq entries in kbest
+filter=uniq             # only unique entries in kbest (surface form)
 pair_sampling=108010    # 10 vs 80 vs 10 and 80 vs 10
-pair_threshold=0        # minimum distance in BLEU
+pair_threshold=0        # minimum distance in BLEU (this will still only use pairs with diff > 0)
 select_weights=last     # just output last weights