diff options
Diffstat (limited to 'training/dtrain/examples/standard')
| -rw-r--r-- | training/dtrain/examples/standard/dtrain.ini | 6 | ||||
| -rw-r--r-- | training/dtrain/examples/standard/expected-output | 115 |
2 files changed, 74 insertions, 47 deletions
diff --git a/training/dtrain/examples/standard/dtrain.ini b/training/dtrain/examples/standard/dtrain.ini index c0912a62..e6d6382e 100644 --- a/training/dtrain/examples/standard/dtrain.ini +++ b/training/dtrain/examples/standard/dtrain.ini @@ -1,6 +1,6 @@ input=./nc-wmt11.de.gz refs=./nc-wmt11.en.gz -output=asdf # a weights file (add .gz for gzip compression) or STDOUT '-' +output=- # a weights file (add .gz for gzip compression) or STDOUT '-' select_weights=VOID # output average (over epochs) weight vector decoder_config=./cdec.ini # config for cdec # weights for these features will be printed on each iteration @@ -10,11 +10,11 @@ print_weights=Glue WordPenalty LanguageModel LanguageModel_OOV PhraseModel_0 Phr stop_after=10 # stop epoch after 10 inputs # interesting stuff -epochs=2 # run over input 2 times +epochs=3 # run over input 3 times k=100 # use 100best lists N=4 # optimize (approx) BLEU4 scorer=fixed_stupid_bleu # use 'stupid' BLEU+1 -learning_rate=1.0 # learning rate, don't care if gamma=0 (perceptron) +learning_rate=1.0 # learning rate, don't care if gamma=0 (perceptron) and loss_margin=0 (not margin perceptron) gamma=0 # use SVM reg sample_from=kbest # use kbest lists (as opposed to forest) filter=uniq # only unique entries in kbest (surface form) diff --git a/training/dtrain/examples/standard/expected-output b/training/dtrain/examples/standard/expected-output index 21f91244..a35bbe6f 100644 --- a/training/dtrain/examples/standard/expected-output +++ b/training/dtrain/examples/standard/expected-output @@ -4,13 +4,13 @@ Reading ./nc-wmt11.en.srilm.gz ----5---10---15---20---25---30---35---40---45---50---55---60---65---70---75---80---85---90---95--100 **************************************************************************************************** Example feature: Shape_S00000_T00000 -Seeding random number sequence to 970626287 +Seeding random number sequence to 4049211323 dtrain Parameters: k 100 N 4 - T 2 + T 3 scorer 'fixed_stupid_bleu' sample from 'kbest' filter 'uniq' @@ -23,6 +23,7 @@ Parameters: pair threshold 0 select weights 'VOID' l1 reg 0 'none' + pclr no max pairs 4294967295 cdec cfg './cdec.ini' input './nc-wmt11.de.gz' @@ -30,62 +31,88 @@ Parameters: output '-' stop_after 10 (a dot represents 10 inputs) -Iteration #1 of 2. +Iteration #1 of 3. . 10 Stopping after 10 input sentences. WEIGHTS - Glue = -614 - WordPenalty = +1256.8 - LanguageModel = +5610.5 - LanguageModel_OOV = -1449 - PhraseModel_0 = -2107 - PhraseModel_1 = -4666.1 - PhraseModel_2 = -2713.5 - PhraseModel_3 = +4204.3 - PhraseModel_4 = -1435.8 - PhraseModel_5 = +916 - PhraseModel_6 = +190 - PassThrough = -2527 + Glue = -1100 + WordPenalty = -82.082 + LanguageModel = -3199.1 + LanguageModel_OOV = -192 + PhraseModel_0 = +3128.2 + PhraseModel_1 = -1610.2 + PhraseModel_2 = -4336.5 + PhraseModel_3 = +2910.3 + PhraseModel_4 = +2523.2 + PhraseModel_5 = +506 + PhraseModel_6 = +1467 + PassThrough = -387 --- - 1best avg score: 0.17874 (+0.17874) - 1best avg model score: 88399 (+88399) - avg # pairs: 798.2 (meaningless) - avg # rank err: 798.2 + 1best avg score: 0.16966 (+0.16966) + 1best avg model score: 2.9874e+05 (+2.9874e+05) + avg # pairs: 906.3 (meaningless) + avg # rank err: 906.3 avg # margin viol: 0 - non0 feature count: 887 + non0 feature count: 825 avg list sz: 91.3 - avg f count: 126.85 -(time 0.33 min, 2 s/S) + avg f count: 139.77 +(time 0.35 min, 2.1 s/S) -Iteration #2 of 2. +Iteration #2 of 3. . 10 WEIGHTS - Glue = -1025 - WordPenalty = +1751.5 - LanguageModel = +10059 - LanguageModel_OOV = -4490 - PhraseModel_0 = -2640.7 - PhraseModel_1 = -3757.4 - PhraseModel_2 = -1133.1 - PhraseModel_3 = +1837.3 - PhraseModel_4 = -3534.3 - PhraseModel_5 = +2308 - PhraseModel_6 = +1677 - PassThrough = -6222 + Glue = -1221 + WordPenalty = +836.89 + LanguageModel = +2332.3 + LanguageModel_OOV = -1451 + PhraseModel_0 = +1507.2 + PhraseModel_1 = -2728.4 + PhraseModel_2 = -4183.6 + PhraseModel_3 = +1816.3 + PhraseModel_4 = -2894.7 + PhraseModel_5 = +1403 + PhraseModel_6 = +35 + PassThrough = -1097 --- - 1best avg score: 0.30764 (+0.12891) - 1best avg model score: -2.5042e+05 (-3.3882e+05) - avg # pairs: 725.9 (meaningless) - avg # rank err: 725.9 + 1best avg score: 0.17399 (+0.004325) + 1best avg model score: 49369 (-2.4937e+05) + avg # pairs: 662.4 (meaningless) + avg # rank err: 662.4 avg # margin viol: 0 - non0 feature count: 1499 + non0 feature count: 1235 avg list sz: 91.3 - avg f count: 114.34 -(time 0.32 min, 1.9 s/S) + avg f count: 125.11 +(time 0.27 min, 1.6 s/S) + +Iteration #3 of 3. + . 10 +WEIGHTS + Glue = -1574 + WordPenalty = -17.372 + LanguageModel = +6861.8 + LanguageModel_OOV = -3997 + PhraseModel_0 = -398.76 + PhraseModel_1 = -3419.6 + PhraseModel_2 = -3186.7 + PhraseModel_3 = +1050.8 + PhraseModel_4 = -2902.7 + PhraseModel_5 = -486 + PhraseModel_6 = -436 + PassThrough = -2985 + --- + 1best avg score: 0.30742 (+0.13343) + 1best avg model score: -1.5393e+05 (-2.0329e+05) + avg # pairs: 623.8 (meaningless) + avg # rank err: 623.8 + avg # margin viol: 0 + non0 feature count: 1770 + avg list sz: 91.3 + avg f count: 118.58 +(time 0.25 min, 1.5 s/S) Writing weights file to '-' ... done --- -Best iteration: 2 [SCORE 'fixed_stupid_bleu'=0.30764]. -This took 0.65 min. +Best iteration: 3 [SCORE 'fixed_stupid_bleu'=0.30742]. +This took 0.86667 min. |