TAU2019-028/preprocess-train/model/train.log

1721 lines
133 KiB
Plaintext

[2020-01-24 02:16:35] [marian] Marian v1.8.33 eba7aed 2019-12-13 11:09:02 -0800
[2020-01-24 02:16:35] [marian] Running on jakub-Lenovo-ideapad-330-15ICH as process 11600 with command line:
[2020-01-24 02:16:35] [marian] /home/jakub/TAU/TAU_24/marian-dev/build/marian --devices 0 --type s2s -e 5 --model preprocess-train/model/model.npz --train-sets preprocess-train/data/europarl.tc.cs preprocess-train/data/europarl.tc.en --vocabs preprocess-train/model/vocab.cs.yml preprocess-train/model/vocab.en.yml --dim-vocabs 50000 50000 --mini-batch-fit -w 1024 --valid-freq 10000 --save-freq 10000 --disp-freq 1000 --valid-mini-batch 23 --valid-max-length 100 --valid-metrics cross-entropy translation --valid-sets preprocess-train/data/valideuroparl.tc.cs preprocess-train/data/valideuroparl.tc.en --valid-script-path 'bash ./preprocess-train/scripts/validate.sh' --log preprocess-train/model/train.log --valid-log preprocess-train/model/valid.log --overwrite --keep-best
[2020-01-24 02:16:35] [config] after-batches: 0
[2020-01-24 02:16:35] [config] after-epochs: 5
[2020-01-24 02:16:35] [config] all-caps-every: 0
[2020-01-24 02:16:35] [config] allow-unk: false
[2020-01-24 02:16:35] [config] authors: false
[2020-01-24 02:16:35] [config] beam-size: 12
[2020-01-24 02:16:35] [config] bert-class-symbol: "[CLS]"
[2020-01-24 02:16:35] [config] bert-mask-symbol: "[MASK]"
[2020-01-24 02:16:35] [config] bert-masking-fraction: 0.15
[2020-01-24 02:16:35] [config] bert-sep-symbol: "[SEP]"
[2020-01-24 02:16:35] [config] bert-train-type-embeddings: true
[2020-01-24 02:16:35] [config] bert-type-vocab-size: 2
[2020-01-24 02:16:35] [config] cite: false
[2020-01-24 02:16:35] [config] clip-gemm: 0
[2020-01-24 02:16:35] [config] clip-norm: 1
[2020-01-24 02:16:35] [config] cost-scaling:
[2020-01-24 02:16:35] [config] []
[2020-01-24 02:16:35] [config] cost-type: ce-mean
[2020-01-24 02:16:35] [config] cpu-threads: 0
[2020-01-24 02:16:35] [config] data-weighting: ""
[2020-01-24 02:16:35] [config] data-weighting-type: sentence
[2020-01-24 02:16:35] [config] dec-cell: gru
[2020-01-24 02:16:35] [config] dec-cell-base-depth: 2
[2020-01-24 02:16:35] [config] dec-cell-high-depth: 1
[2020-01-24 02:16:35] [config] dec-depth: 1
[2020-01-24 02:16:35] [config] devices:
[2020-01-24 02:16:35] [config] - 0
[2020-01-24 02:16:35] [config] dim-emb: 512
[2020-01-24 02:16:35] [config] dim-rnn: 1024
[2020-01-24 02:16:35] [config] dim-vocabs:
[2020-01-24 02:16:35] [config] - 50000
[2020-01-24 02:16:35] [config] - 50000
[2020-01-24 02:16:35] [config] disp-first: 0
[2020-01-24 02:16:35] [config] disp-freq: 1000
[2020-01-24 02:16:35] [config] disp-label-counts: false
[2020-01-24 02:16:35] [config] dropout-rnn: 0
[2020-01-24 02:16:35] [config] dropout-src: 0
[2020-01-24 02:16:35] [config] dropout-trg: 0
[2020-01-24 02:16:35] [config] dump-config: ""
[2020-01-24 02:16:35] [config] early-stopping: 10
[2020-01-24 02:16:35] [config] embedding-fix-src: false
[2020-01-24 02:16:35] [config] embedding-fix-trg: false
[2020-01-24 02:16:35] [config] embedding-normalization: false
[2020-01-24 02:16:35] [config] embedding-vectors:
[2020-01-24 02:16:35] [config] []
[2020-01-24 02:16:35] [config] enc-cell: gru
[2020-01-24 02:16:35] [config] enc-cell-depth: 1
[2020-01-24 02:16:35] [config] enc-depth: 1
[2020-01-24 02:16:35] [config] enc-type: bidirectional
[2020-01-24 02:16:35] [config] english-title-case-every: 0
[2020-01-24 02:16:35] [config] exponential-smoothing: 0
[2020-01-24 02:16:35] [config] factor-weight: 1
[2020-01-24 02:16:35] [config] grad-dropping-momentum: 0
[2020-01-24 02:16:35] [config] grad-dropping-rate: 0
[2020-01-24 02:16:35] [config] grad-dropping-warmup: 100
[2020-01-24 02:16:35] [config] gradient-checkpointing: false
[2020-01-24 02:16:35] [config] guided-alignment: none
[2020-01-24 02:16:35] [config] guided-alignment-cost: mse
[2020-01-24 02:16:35] [config] guided-alignment-weight: 0.1
[2020-01-24 02:16:35] [config] ignore-model-config: false
[2020-01-24 02:16:35] [config] input-types:
[2020-01-24 02:16:35] [config] []
[2020-01-24 02:16:35] [config] interpolate-env-vars: false
[2020-01-24 02:16:35] [config] keep-best: true
[2020-01-24 02:16:35] [config] label-smoothing: 0
[2020-01-24 02:16:35] [config] layer-normalization: false
[2020-01-24 02:16:35] [config] learn-rate: 0.0001
[2020-01-24 02:16:35] [config] lemma-dim-emb: 0
[2020-01-24 02:16:35] [config] log: preprocess-train/model/train.log
[2020-01-24 02:16:35] [config] log-level: info
[2020-01-24 02:16:35] [config] log-time-zone: ""
[2020-01-24 02:16:35] [config] lr-decay: 0
[2020-01-24 02:16:35] [config] lr-decay-freq: 50000
[2020-01-24 02:16:35] [config] lr-decay-inv-sqrt:
[2020-01-24 02:16:35] [config] - 0
[2020-01-24 02:16:35] [config] lr-decay-repeat-warmup: false
[2020-01-24 02:16:35] [config] lr-decay-reset-optimizer: false
[2020-01-24 02:16:35] [config] lr-decay-start:
[2020-01-24 02:16:35] [config] - 10
[2020-01-24 02:16:35] [config] - 1
[2020-01-24 02:16:35] [config] lr-decay-strategy: epoch+stalled
[2020-01-24 02:16:35] [config] lr-report: false
[2020-01-24 02:16:35] [config] lr-warmup: 0
[2020-01-24 02:16:35] [config] lr-warmup-at-reload: false
[2020-01-24 02:16:35] [config] lr-warmup-cycle: false
[2020-01-24 02:16:35] [config] lr-warmup-start-rate: 0
[2020-01-24 02:16:35] [config] max-length: 50
[2020-01-24 02:16:35] [config] max-length-crop: false
[2020-01-24 02:16:35] [config] max-length-factor: 3
[2020-01-24 02:16:35] [config] maxi-batch: 100
[2020-01-24 02:16:35] [config] maxi-batch-sort: trg
[2020-01-24 02:16:35] [config] mini-batch: 64
[2020-01-24 02:16:35] [config] mini-batch-fit: true
[2020-01-24 02:16:35] [config] mini-batch-fit-step: 10
[2020-01-24 02:16:35] [config] mini-batch-overstuff: 1
[2020-01-24 02:16:35] [config] mini-batch-track-lr: false
[2020-01-24 02:16:35] [config] mini-batch-understuff: 1
[2020-01-24 02:16:35] [config] mini-batch-warmup: 0
[2020-01-24 02:16:35] [config] mini-batch-words: 0
[2020-01-24 02:16:35] [config] mini-batch-words-ref: 0
[2020-01-24 02:16:35] [config] model: preprocess-train/model/model.npz
[2020-01-24 02:16:35] [config] multi-loss-type: sum
[2020-01-24 02:16:35] [config] multi-node: false
[2020-01-24 02:16:35] [config] multi-node-overlap: true
[2020-01-24 02:16:35] [config] n-best: false
[2020-01-24 02:16:35] [config] no-nccl: false
[2020-01-24 02:16:35] [config] no-reload: false
[2020-01-24 02:16:35] [config] no-restore-corpus: false
[2020-01-24 02:16:35] [config] normalize: 0
[2020-01-24 02:16:35] [config] normalize-gradient: false
[2020-01-24 02:16:35] [config] num-devices: 0
[2020-01-24 02:16:35] [config] optimizer: adam
[2020-01-24 02:16:35] [config] optimizer-delay: 1
[2020-01-24 02:16:35] [config] optimizer-params:
[2020-01-24 02:16:35] [config] []
[2020-01-24 02:16:35] [config] overwrite: true
[2020-01-24 02:16:35] [config] precision:
[2020-01-24 02:16:35] [config] - float32
[2020-01-24 02:16:35] [config] - float32
[2020-01-24 02:16:35] [config] - float32
[2020-01-24 02:16:35] [config] pretrained-model: ""
[2020-01-24 02:16:35] [config] quiet: false
[2020-01-24 02:16:35] [config] quiet-translation: false
[2020-01-24 02:16:35] [config] relative-paths: false
[2020-01-24 02:16:35] [config] right-left: false
[2020-01-24 02:16:35] [config] save-freq: 10000
[2020-01-24 02:16:35] [config] seed: 0
[2020-01-24 02:16:35] [config] shuffle: data
[2020-01-24 02:16:35] [config] shuffle-in-ram: false
[2020-01-24 02:16:35] [config] skip: false
[2020-01-24 02:16:35] [config] sqlite: ""
[2020-01-24 02:16:35] [config] sqlite-drop: false
[2020-01-24 02:16:35] [config] sync-sgd: false
[2020-01-24 02:16:35] [config] tempdir: /tmp
[2020-01-24 02:16:35] [config] tied-embeddings: false
[2020-01-24 02:16:35] [config] tied-embeddings-all: false
[2020-01-24 02:16:35] [config] tied-embeddings-src: false
[2020-01-24 02:16:35] [config] train-sets:
[2020-01-24 02:16:35] [config] - preprocess-train/data/europarl.tc.cs
[2020-01-24 02:16:35] [config] - preprocess-train/data/europarl.tc.en
[2020-01-24 02:16:35] [config] transformer-aan-activation: swish
[2020-01-24 02:16:35] [config] transformer-aan-depth: 2
[2020-01-24 02:16:35] [config] transformer-aan-nogate: false
[2020-01-24 02:16:35] [config] transformer-decoder-autoreg: self-attention
[2020-01-24 02:16:35] [config] transformer-depth-scaling: false
[2020-01-24 02:16:35] [config] transformer-dim-aan: 2048
[2020-01-24 02:16:35] [config] transformer-dim-ffn: 2048
[2020-01-24 02:16:35] [config] transformer-dropout: 0
[2020-01-24 02:16:35] [config] transformer-dropout-attention: 0
[2020-01-24 02:16:35] [config] transformer-dropout-ffn: 0
[2020-01-24 02:16:35] [config] transformer-ffn-activation: swish
[2020-01-24 02:16:35] [config] transformer-ffn-depth: 2
[2020-01-24 02:16:35] [config] transformer-guided-alignment-layer: last
[2020-01-24 02:16:35] [config] transformer-heads: 8
[2020-01-24 02:16:35] [config] transformer-no-projection: false
[2020-01-24 02:16:35] [config] transformer-postprocess: dan
[2020-01-24 02:16:35] [config] transformer-postprocess-emb: d
[2020-01-24 02:16:35] [config] transformer-preprocess: ""
[2020-01-24 02:16:35] [config] transformer-tied-layers:
[2020-01-24 02:16:35] [config] []
[2020-01-24 02:16:35] [config] transformer-train-position-embeddings: false
[2020-01-24 02:16:35] [config] type: s2s
[2020-01-24 02:16:35] [config] ulr: false
[2020-01-24 02:16:35] [config] ulr-dim-emb: 0
[2020-01-24 02:16:35] [config] ulr-dropout: 0
[2020-01-24 02:16:35] [config] ulr-keys-vectors: ""
[2020-01-24 02:16:35] [config] ulr-query-vectors: ""
[2020-01-24 02:16:35] [config] ulr-softmax-temperature: 1
[2020-01-24 02:16:35] [config] ulr-trainable-transformation: false
[2020-01-24 02:16:35] [config] unlikelihood-loss: false
[2020-01-24 02:16:35] [config] valid-freq: 10000
[2020-01-24 02:16:35] [config] valid-log: preprocess-train/model/valid.log
[2020-01-24 02:16:35] [config] valid-max-length: 100
[2020-01-24 02:16:35] [config] valid-metrics:
[2020-01-24 02:16:35] [config] - cross-entropy
[2020-01-24 02:16:35] [config] - translation
[2020-01-24 02:16:35] [config] valid-mini-batch: 23
[2020-01-24 02:16:35] [config] valid-script-args:
[2020-01-24 02:16:35] [config] []
[2020-01-24 02:16:35] [config] valid-script-path: bash ./preprocess-train/scripts/validate.sh
[2020-01-24 02:16:35] [config] valid-sets:
[2020-01-24 02:16:35] [config] - preprocess-train/data/valideuroparl.tc.cs
[2020-01-24 02:16:35] [config] - preprocess-train/data/valideuroparl.tc.en
[2020-01-24 02:16:35] [config] valid-translation-output: ""
[2020-01-24 02:16:35] [config] vocabs:
[2020-01-24 02:16:35] [config] - preprocess-train/model/vocab.cs.yml
[2020-01-24 02:16:35] [config] - preprocess-train/model/vocab.en.yml
[2020-01-24 02:16:35] [config] word-penalty: 0
[2020-01-24 02:16:35] [config] workspace: 1024
[2020-01-24 02:16:35] [config] Model is being created with Marian v1.8.33 eba7aed 2019-12-13 11:09:02 -0800
[2020-01-24 02:16:35] Using single-device training
[2020-01-24 02:16:35] [data] Loading vocabulary from JSON/Yaml file preprocess-train/model/vocab.cs.yml
[2020-01-24 02:16:35] [data] Setting vocabulary size for input 0 to 50000
[2020-01-24 02:16:35] [data] Loading vocabulary from JSON/Yaml file preprocess-train/model/vocab.en.yml
[2020-01-24 02:16:36] [data] Setting vocabulary size for input 1 to 50000
[2020-01-24 02:16:36] Compiled without MPI support. Falling back to FakeMPIWrapper
[2020-01-24 02:16:36] [batching] Collecting statistics for batch fitting with step size 10
[2020-01-24 02:16:36] [memory] Extending reserved space to 1024 MB (device gpu0)
[2020-01-24 02:16:36] [logits] applyLossFunction() for 1 factors
[2020-01-24 02:16:36] [memory] Reserving 422 MB, device gpu0
[2020-01-24 02:16:36] [gpu] 16-bit TensorCores enabled for float32 matrix operations
[2020-01-24 02:16:37] [memory] Reserving 422 MB, device gpu0
[2020-01-24 02:16:44] [batching] Done. Typical MB size is 1562 target words
[2020-01-24 02:16:44] [memory] Extending reserved space to 1024 MB (device gpu0)
[2020-01-24 02:16:44] Training started
[2020-01-24 02:16:44] [data] Shuffling data
[2020-01-24 02:16:44] [data] Done reading 640633 sentences
[2020-01-24 02:16:47] [data] Done shuffling 640633 sentences to temp files
[2020-01-24 02:16:47] [memory] Reserving 422 MB, device gpu0
[2020-01-24 02:16:48] [memory] Reserving 422 MB, device gpu0
[2020-01-24 02:16:48] [memory] Reserving 844 MB, device gpu0
[2020-01-24 02:26:26] Ep. 1 : Up. 1000 : Sen. 48,745 : Cost 151.28079224 : Time 590.95s : 2052.93 words/s
[2020-01-24 02:36:09] Ep. 1 : Up. 2000 : Sen. 97,401 : Cost 126.55426025 : Time 582.81s : 2078.21 words/s
[2020-01-24 02:45:52] Ep. 1 : Up. 3000 : Sen. 145,828 : Cost 115.34564209 : Time 582.88s : 2067.70 words/s
[2020-01-24 02:55:37] Ep. 1 : Up. 4000 : Sen. 194,574 : Cost 106.52265930 : Time 585.27s : 2073.09 words/s
[2020-01-24 03:05:16] Ep. 1 : Up. 5000 : Sen. 243,200 : Cost 96.68421173 : Time 578.35s : 2090.16 words/s
[2020-01-24 03:14:54] Ep. 1 : Up. 6000 : Sen. 291,887 : Cost 88.26989746 : Time 578.69s : 2100.87 words/s
[2020-01-24 03:24:31] Ep. 1 : Up. 7000 : Sen. 340,686 : Cost 80.46852112 : Time 576.92s : 2097.51 words/s
[2020-01-24 03:34:08] Ep. 1 : Up. 8000 : Sen. 389,260 : Cost 74.79323578 : Time 576.35s : 2097.24 words/s
[2020-01-24 03:43:45] Ep. 1 : Up. 9000 : Sen. 437,816 : Cost 70.13729858 : Time 577.15s : 2101.88 words/s
[2020-01-24 03:53:22] Ep. 1 : Up. 10000 : Sen. 486,859 : Cost 65.71810913 : Time 577.51s : 2102.19 words/s
[2020-01-24 03:53:37] Saving model weights and runtime parameters to preprocess-train/model/model.npz.best-cross-entropy.npz
[2020-01-24 03:53:38] [valid] Ep. 1 : Up. 10000 : cross-entropy : 108.847 : new best
[2020-01-24 03:53:39] Translating validation set...
[2020-01-24 03:53:51] Best translation 0 : the strategy is in favour of the agenda
[2020-01-24 03:53:51] Best translation 1 : there is a policy that we have a policy for the fight against the Hungarian fraud .
[2020-01-24 03:53:51] Best translation 2 : however , the United Kingdom , of course , is that , in the UK , it is less than in the US , less than the number of people that have been made by the number of people .
[2020-01-24 03:53:51] Best translation 3 : at the last few years , however , there are only 300 cases in the US in the US .
[2020-01-24 03:53:51] Best translation 4 : one thing is certain : this new provisions will have a negative impact on recent participation .
[2020-01-24 03:53:51] Best translation 5 : in this sense , these measures are at the heart of the US democratic system .
[2020-01-24 03:53:51] Best translation 10 : the new legislative law require the voters of the Czech Republic of Justice and the fact about the US authorities .
[2020-01-24 03:53:51] Best translation 20 : at the same time , there are often more than five times .
[2020-01-24 03:53:52] Best translation 40 : on the other hand , let the other hand , who did not cut up to a few % of all of the voters , as a result of 90 % of the voters who have spoken in it .
[2020-01-24 03:53:52] Best translation 80 : 13 % of cases are in place .
[2020-01-24 03:53:52] Best translation 160 : ' in the lives of the lives , it has shown that we do not get up to the streets of Ireland , the ' lack of Mugabe , in particular , in @-@ depth .
[2020-01-24 03:54:03] Best translation 320 : this means that it is a great deal , and that it gives a great deal .
[2020-01-24 03:54:17] Best translation 640 : in relation to Russia 's proposal , the proposal to defend the law in relation to the removal of civilian weapons .
[2020-01-24 03:55:01] Best translation 1280 : finally , as regards the current society , the current society is based on the recognition of their own citizens , regardless of the most important interest .
[2020-01-24 03:56:21] Best translation 2560 : after all , what has been a great ban on the ban which , together with 70 times , are not going to be seen in the two of the two @-@ Saharan @-@ Saharan @-@ Qaeda .
[2020-01-24 03:56:39] Total translation time: 180.33303s
[2020-01-24 03:56:40] Saving model weights and runtime parameters to preprocess-train/model/model.npz.best-translation.npz
[2020-01-24 03:56:42] [valid] Ep. 1 : Up. 10000 : translation : 7.52 : new best
[2020-01-24 03:56:42] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-24 03:56:43] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-24 04:06:29] Ep. 1 : Up. 11000 : Sen. 535,414 : Cost 63.08740997 : Time 786.64s : 1538.94 words/s
[2020-01-24 04:16:05] Ep. 1 : Up. 12000 : Sen. 583,807 : Cost 60.32462692 : Time 575.46s : 2097.88 words/s
[2020-01-24 04:17:16] Seen 589960 samples
[2020-01-24 04:17:16] Starting epoch 2
[2020-01-24 04:17:16] [data] Shuffling data
[2020-01-24 04:17:16] [data] Done reading 640633 sentences
[2020-01-24 04:17:19] [data] Done shuffling 640633 sentences to temp files
[2020-01-24 04:25:46] Ep. 2 : Up. 13000 : Sen. 42,890 : Cost 56.30655289 : Time 581.98s : 2082.81 words/s
[2020-01-24 04:35:25] Ep. 2 : Up. 14000 : Sen. 91,479 : Cost 55.08889771 : Time 578.02s : 2096.13 words/s
[2020-01-24 04:45:02] Ep. 2 : Up. 15000 : Sen. 140,275 : Cost 53.31164932 : Time 577.25s : 2098.27 words/s
[2020-01-24 04:54:42] Ep. 2 : Up. 16000 : Sen. 189,016 : Cost 52.27413940 : Time 579.79s : 2098.26 words/s
[2020-01-24 05:04:17] Ep. 2 : Up. 17000 : Sen. 237,393 : Cost 51.21431732 : Time 575.18s : 2100.76 words/s
[2020-01-24 05:13:53] Ep. 2 : Up. 18000 : Sen. 286,096 : Cost 49.68261337 : Time 575.93s : 2099.53 words/s
[2020-01-24 05:23:30] Ep. 2 : Up. 19000 : Sen. 335,061 : Cost 48.82456970 : Time 576.83s : 2102.60 words/s
[2020-01-24 05:33:06] Ep. 2 : Up. 20000 : Sen. 383,511 : Cost 48.51165771 : Time 576.38s : 2099.59 words/s
[2020-01-24 05:33:20] Saving model weights and runtime parameters to preprocess-train/model/model.npz.best-cross-entropy.npz
[2020-01-24 05:33:26] [valid] Ep. 2 : Up. 20000 : cross-entropy : 94.6385 : new best
[2020-01-24 05:33:26] Translating validation set...
[2020-01-24 05:33:38] Best translation 0 : strategy on the Left
[2020-01-24 05:33:38] Best translation 1 : the leaders of the leaders have not taken its policy to combat electoral fraud .
[2020-01-24 05:33:38] Best translation 2 : however , the centre of the Centre says such a scandal that electoral fraud is less difficult in the US than the number of people killed after the intervention .
[2020-01-24 05:33:38] Best translation 3 : for the last ten years , after all , there are only 300 cases of electoral fraud in the US .
[2020-01-24 05:33:38] Best translation 4 : one is certain : this new provisions will have negative effects on electoral participation .
[2020-01-24 05:33:38] Best translation 5 : in this sense , this measure is partly by the American democratic system .
[2020-01-24 05:33:38] Best translation 10 : the new electoral laws require voters to reject the identity of the identity identity and the government of American citizenship .
[2020-01-24 05:33:38] Best translation 20 : in addition , there are often more than 100 % of the ability of identity .
[2020-01-24 05:33:38] Best translation 40 : on the other hand , I do not want to see 76 % of all voters , only 46 % of the voters who voted in favour .
[2020-01-24 05:33:38] Best translation 80 : cigarettes are to 85 % of cases of cancer .
[2020-01-24 05:33:38] Best translation 160 : ' in a daily life is that you are afraid that we are going to have an anti @-@ year camp , ' it is a sine Ferrero @-@ moon , the High Representative .
[2020-01-24 05:33:48] Best translation 320 : this means that it is very integrated and that it gives a great deal .
[2020-01-24 05:34:01] Best translation 640 : in writing .
[2020-01-24 05:34:41] Best translation 1280 : finally , as a result of the current society , the current company has been based on the ill @-@ out of their own benefit , regardless of the interest interest .
[2020-01-24 05:35:58] Best translation 2560 : after all , what has been created by a large percentage of the waste which , together with the 70 years , is to blame them at the beginning of a two @-@ year @-@ off @-@ old @-@ off @-@ old @-@ off @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@ old @-@
[2020-01-24 05:36:15] Total translation time: 169.70254s
[2020-01-24 05:36:16] Saving model weights and runtime parameters to preprocess-train/model/model.npz.best-translation.npz
[2020-01-24 05:36:21] [valid] Ep. 2 : Up. 20000 : translation : 12.28 : new best
[2020-01-24 05:36:21] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-24 05:36:26] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-24 05:46:15] Ep. 2 : Up. 21000 : Sen. 432,076 : Cost 47.54589844 : Time 789.34s : 1534.76 words/s
[2020-01-24 05:55:54] Ep. 2 : Up. 22000 : Sen. 480,811 : Cost 46.49859619 : Time 579.22s : 2088.69 words/s
[2020-01-24 06:05:35] Ep. 2 : Up. 23000 : Sen. 529,294 : Cost 46.13855362 : Time 580.15s : 2090.43 words/s
[2020-01-24 06:15:14] Ep. 2 : Up. 24000 : Sen. 578,330 : Cost 45.10533524 : Time 579.56s : 2094.68 words/s
[2020-01-24 06:17:33] Seen 589960 samples
[2020-01-24 06:17:33] Starting epoch 3
[2020-01-24 06:17:33] [data] Shuffling data
[2020-01-24 06:17:33] [data] Done reading 640633 sentences
[2020-01-24 06:17:36] [data] Done shuffling 640633 sentences to temp files
[2020-01-24 06:24:55] Ep. 3 : Up. 25000 : Sen. 36,956 : Cost 42.75318909 : Time 580.39s : 2076.11 words/s
[2020-01-24 06:34:36] Ep. 3 : Up. 26000 : Sen. 85,927 : Cost 41.69101715 : Time 581.44s : 2089.96 words/s
[2020-01-24 06:44:14] Ep. 3 : Up. 27000 : Sen. 134,121 : Cost 42.14825058 : Time 577.98s : 2090.04 words/s
[2020-01-24 06:53:47] Ep. 3 : Up. 28000 : Sen. 182,537 : Cost 41.62125397 : Time 573.27s : 2099.74 words/s
[2020-01-24 07:03:24] Ep. 3 : Up. 29000 : Sen. 231,166 : Cost 41.35540390 : Time 576.41s : 2100.25 words/s
[2020-01-24 07:13:00] Ep. 3 : Up. 30000 : Sen. 280,005 : Cost 40.73836136 : Time 576.58s : 2104.21 words/s
[2020-01-24 07:13:15] Saving model weights and runtime parameters to preprocess-train/model/model.npz.best-cross-entropy.npz
[2020-01-24 07:13:19] [valid] Ep. 3 : Up. 30000 : cross-entropy : 89.3 : new best
[2020-01-24 07:13:19] Translating validation set...
[2020-01-24 07:13:31] Best translation 0 : the strategy of the visitors '
[2020-01-24 07:13:31] Best translation 1 : the Burmese leaders did not give its policy a viable fight against electoral fraud .
[2020-01-24 07:13:31] Best translation 2 : however , the Sixth Centre is described as a myth that electoral fraud is less frequent in the USA than the number of people who die after intervention .
[2020-01-24 07:13:31] Best translation 3 : after all , there was only 300 cases of electoral fraud in the USA over the last ten years .
[2020-01-24 07:13:31] Best translation 4 : one is certain : these new provisions will have a negative impact on electoral participation .
[2020-01-24 07:13:31] Best translation 5 : in this sense , this measure has partly exacerbated the American democratic system .
[2020-01-24 07:13:31] Best translation 10 : new electoral laws require the voters to submit a criminal identity card and proof of the American State citizenship .
[2020-01-24 07:13:31] Best translation 20 : in addition , they are often more than a hundred dollars on identity card .
[2020-01-24 07:13:31] Best translation 40 : on the other hand , those who would like 76 % of all voters would like to see 46 % of the voters who voted in advance .
[2020-01-24 07:13:31] Best translation 80 : cigarettes are worth 85 % of cancer cases .
[2020-01-24 07:13:31] Best translation 160 : where there is a lot of people who have been able to learn from the fact that the six @-@ year @-@ based version of Sarajevo , the High Representative , was set up .
[2020-01-24 07:13:41] Best translation 320 : this means that it is a very blessing , and that it gives a great deal .
[2020-01-24 07:13:52] Best translation 640 : in writing .
[2020-01-24 07:14:30] Best translation 1280 : in writing . - ( DE ) The conclusion is that the current company is based on the brutal advantage of their own benefit , regardless of the emotional interest .
[2020-01-24 07:15:44] Best translation 2560 : after all , what about the great blocks of sulphur , which we attach to 70 times together , is at the mercy of a two @-@ year @-@ old @-@ old shop .
[2020-01-24 07:16:00] Total translation time: 160.62914s
[2020-01-24 07:16:01] Saving model weights and runtime parameters to preprocess-train/model/model.npz.best-translation.npz
[2020-01-24 07:16:05] [valid] Ep. 3 : Up. 30000 : translation : 14.42 : new best
[2020-01-24 07:16:05] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-24 07:16:10] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-24 07:26:02] Ep. 3 : Up. 31000 : Sen. 328,656 : Cost 40.77697372 : Time 781.78s : 1554.38 words/s
[2020-01-24 07:35:38] Ep. 3 : Up. 32000 : Sen. 377,471 : Cost 40.20654297 : Time 576.46s : 2097.85 words/s
[2020-01-24 07:45:14] Ep. 3 : Up. 33000 : Sen. 425,938 : Cost 40.16862488 : Time 575.40s : 2099.89 words/s
[2020-01-24 07:54:50] Ep. 3 : Up. 34000 : Sen. 474,494 : Cost 40.06607819 : Time 576.33s : 2100.49 words/s
[2020-01-24 08:04:25] Ep. 3 : Up. 35000 : Sen. 522,935 : Cost 39.58546066 : Time 574.88s : 2096.76 words/s
[2020-01-24 08:14:03] Ep. 3 : Up. 36000 : Sen. 571,658 : Cost 39.56554031 : Time 577.45s : 2101.92 words/s
[2020-01-24 08:17:38] Seen 589960 samples
[2020-01-24 08:17:38] Starting epoch 4
[2020-01-24 08:17:38] [data] Shuffling data
[2020-01-24 08:17:38] [data] Done reading 640633 sentences
[2020-01-24 08:17:41] [data] Done shuffling 640633 sentences to temp files
[2020-01-24 08:23:42] Ep. 4 : Up. 37000 : Sen. 30,677 : Cost 37.15330505 : Time 579.74s : 2086.84 words/s
[2020-01-24 08:33:16] Ep. 4 : Up. 38000 : Sen. 79,075 : Cost 36.37662888 : Time 573.81s : 2096.89 words/s
[2020-01-24 08:42:53] Ep. 4 : Up. 39000 : Sen. 127,595 : Cost 36.62038803 : Time 576.97s : 2098.38 words/s
[2020-01-24 08:52:33] Ep. 4 : Up. 40000 : Sen. 176,327 : Cost 36.57798386 : Time 579.80s : 2100.08 words/s
[2020-01-24 08:52:47] Saving model weights and runtime parameters to preprocess-train/model/model.npz.best-cross-entropy.npz
[2020-01-24 08:52:52] [valid] Ep. 4 : Up. 40000 : cross-entropy : 86.3268 : new best
[2020-01-24 08:52:53] Translating validation set...
[2020-01-24 08:53:03] Best translation 0 : the Danube Strategy against New Zealand
[2020-01-24 08:53:03] Best translation 1 : there is a need for the fight against electoral fraud .
[2020-01-24 08:53:03] Best translation 2 : however , the centre @-@ based centre is the myth of the myth that electoral fraud in the US is less well than the number of people who die after intervention .
[2020-01-24 08:53:03] Best translation 3 : after all , it was only 300 cases of election fraud in the US over the last ten years .
[2020-01-24 08:53:03] Best translation 4 : one thing is certain : these new provisions will have a negative impact on electoral participation .
[2020-01-24 08:53:03] Best translation 5 : in this sense , this measure has partly exacerbated the US democratic system .
[2020-01-24 08:53:03] Best translation 10 : the new electoral laws require voters to submit the identity card with photographs and proof of the American State citizenship .
[2020-01-24 08:53:03] Best translation 20 : in addition , the purchase required identity card costs more than 100 dollars .
[2020-01-24 08:53:04] Best translation 40 : on the other hand , there were only 46 % of the voters who voted in advance of 76 % of all voters .
[2020-01-24 08:53:04] Best translation 80 : cigarettes stands for 85 % of lung cancer .
[2020-01-24 08:53:04] Best translation 160 : ' around life , you learn that we are not going through a visit , the ' Small Pierre Facility , the High Representative , the High Representative .
[2020-01-24 08:53:13] Best translation 320 : this means that there is a great deal of ambition and it gives a lot to do .
[2020-01-24 08:53:24] Best translation 640 : we are presenting Russia 's proposal to tighten the law in relation to the owners of civil weapons .
[2020-01-24 08:53:59] Best translation 1280 : in conclusion , the current company is based on unscrupulous self @-@ benefit , regardless of the noble interest .
[2020-01-24 08:55:07] Best translation 2560 : after the very large sulphur blocks , which bring together up to 70 times together , is being held at the same time in two stages of a thousand @-@ up shop .
[2020-01-24 08:55:22] Total translation time: 149.11261s
[2020-01-24 08:55:23] Saving model weights and runtime parameters to preprocess-train/model/model.npz.best-translation.npz
[2020-01-24 08:55:27] [valid] Ep. 4 : Up. 40000 : translation : 15.55 : new best
[2020-01-24 08:55:27] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-24 08:55:32] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-24 09:05:52] Ep. 4 : Up. 41000 : Sen. 225,356 : Cost 36.22032547 : Time 798.86s : 1527.07 words/s
[2020-01-24 09:16:49] Ep. 4 : Up. 42000 : Sen. 274,385 : Cost 36.03122711 : Time 657.12s : 1844.75 words/s
[2020-01-24 09:31:13] Ep. 4 : Up. 43000 : Sen. 322,628 : Cost 36.29774475 : Time 864.62s : 1392.28 words/s
[2020-01-24 10:38:28] [marian] Marian v1.8.33 eba7aed 2019-12-13 11:09:02 -0800
[2020-01-24 10:38:28] [marian] Running on jakub-Lenovo-ideapad-330-15ICH as process 2450 with command line:
[2020-01-24 10:38:28] [marian] /home/jakub/TAU/TAU_24/marian-dev/build/marian --devices 0 --type s2s -e 5 --model preprocess-train/model/model.npz --train-sets preprocess-train/data/europarl.tc.cs preprocess-train/data/europarl.tc.en --vocabs preprocess-train/model/vocab.cs.yml preprocess-train/model/vocab.en.yml --dim-vocabs 50000 50000 --mini-batch-fit -w 1024 --valid-freq 10000 --save-freq 10000 --disp-freq 1000 --valid-mini-batch 23 --valid-max-length 100 --valid-metrics cross-entropy translation --valid-sets preprocess-train/data/valideuroparl.tc.cs preprocess-train/data/valideuroparl.tc.en --valid-script-path 'bash ./preprocess-train/scripts/validate.sh' --log preprocess-train/model/train.log --valid-log preprocess-train/model/valid.log --overwrite --keep-best
[2020-01-24 10:38:28] [config] after-batches: 0
[2020-01-24 10:38:28] [config] after-epochs: 5
[2020-01-24 10:38:28] [config] all-caps-every: 0
[2020-01-24 10:38:28] [config] allow-unk: false
[2020-01-24 10:38:28] [config] authors: false
[2020-01-24 10:38:28] [config] beam-size: 12
[2020-01-24 10:38:28] [config] bert-class-symbol: "[CLS]"
[2020-01-24 10:38:28] [config] bert-mask-symbol: "[MASK]"
[2020-01-24 10:38:28] [config] bert-masking-fraction: 0.15
[2020-01-24 10:38:28] [config] bert-sep-symbol: "[SEP]"
[2020-01-24 10:38:28] [config] bert-train-type-embeddings: true
[2020-01-24 10:38:28] [config] bert-type-vocab-size: 2
[2020-01-24 10:38:28] [config] cite: false
[2020-01-24 10:38:28] [config] clip-gemm: 0
[2020-01-24 10:38:28] [config] clip-norm: 1
[2020-01-24 10:38:28] [config] cost-scaling:
[2020-01-24 10:38:28] [config] []
[2020-01-24 10:38:28] [config] cost-type: ce-mean
[2020-01-24 10:38:28] [config] cpu-threads: 0
[2020-01-24 10:38:28] [config] data-weighting: ""
[2020-01-24 10:38:28] [config] data-weighting-type: sentence
[2020-01-24 10:38:28] [config] dec-cell: gru
[2020-01-24 10:38:28] [config] dec-cell-base-depth: 2
[2020-01-24 10:38:28] [config] dec-cell-high-depth: 1
[2020-01-24 10:38:28] [config] dec-depth: 1
[2020-01-24 10:38:28] [config] devices:
[2020-01-24 10:38:28] [config] - 0
[2020-01-24 10:38:28] [config] dim-emb: 512
[2020-01-24 10:38:28] [config] dim-rnn: 1024
[2020-01-24 10:38:28] [config] dim-vocabs:
[2020-01-24 10:38:28] [config] - 50000
[2020-01-24 10:38:28] [config] - 50000
[2020-01-24 10:38:28] [config] disp-first: 0
[2020-01-24 10:38:28] [config] disp-freq: 1000
[2020-01-24 10:38:28] [config] disp-label-counts: false
[2020-01-24 10:38:28] [config] dropout-rnn: 0
[2020-01-24 10:38:28] [config] dropout-src: 0
[2020-01-24 10:38:28] [config] dropout-trg: 0
[2020-01-24 10:38:28] [config] dump-config: ""
[2020-01-24 10:38:28] [config] early-stopping: 10
[2020-01-24 10:38:28] [config] embedding-fix-src: false
[2020-01-24 10:38:28] [config] embedding-fix-trg: false
[2020-01-24 10:38:28] [config] embedding-normalization: false
[2020-01-24 10:38:28] [config] embedding-vectors:
[2020-01-24 10:38:28] [config] []
[2020-01-24 10:38:28] [config] enc-cell: gru
[2020-01-24 10:38:28] [config] enc-cell-depth: 1
[2020-01-24 10:38:28] [config] enc-depth: 1
[2020-01-24 10:38:28] [config] enc-type: bidirectional
[2020-01-24 10:38:28] [config] english-title-case-every: 0
[2020-01-24 10:38:28] [config] exponential-smoothing: 0
[2020-01-24 10:38:28] [config] factor-weight: 1
[2020-01-24 10:38:28] [config] grad-dropping-momentum: 0
[2020-01-24 10:38:28] [config] grad-dropping-rate: 0
[2020-01-24 10:38:28] [config] grad-dropping-warmup: 100
[2020-01-24 10:38:28] [config] gradient-checkpointing: false
[2020-01-24 10:38:28] [config] guided-alignment: none
[2020-01-24 10:38:28] [config] guided-alignment-cost: mse
[2020-01-24 10:38:28] [config] guided-alignment-weight: 0.1
[2020-01-24 10:38:28] [config] ignore-model-config: false
[2020-01-24 10:38:28] [config] input-types:
[2020-01-24 10:38:28] [config] []
[2020-01-24 10:38:28] [config] interpolate-env-vars: false
[2020-01-24 10:38:28] [config] keep-best: true
[2020-01-24 10:38:28] [config] label-smoothing: 0
[2020-01-24 10:38:28] [config] layer-normalization: false
[2020-01-24 10:38:28] [config] learn-rate: 0.0001
[2020-01-24 10:38:28] [config] lemma-dim-emb: 0
[2020-01-24 10:38:28] [config] log: preprocess-train/model/train.log
[2020-01-24 10:38:28] [config] log-level: info
[2020-01-24 10:38:28] [config] log-time-zone: ""
[2020-01-24 10:38:28] [config] lr-decay: 0
[2020-01-24 10:38:28] [config] lr-decay-freq: 50000
[2020-01-24 10:38:28] [config] lr-decay-inv-sqrt:
[2020-01-24 10:38:28] [config] - 0
[2020-01-24 10:38:28] [config] lr-decay-repeat-warmup: false
[2020-01-24 10:38:28] [config] lr-decay-reset-optimizer: false
[2020-01-24 10:38:28] [config] lr-decay-start:
[2020-01-24 10:38:28] [config] - 10
[2020-01-24 10:38:28] [config] - 1
[2020-01-24 10:38:28] [config] lr-decay-strategy: epoch+stalled
[2020-01-24 10:38:28] [config] lr-report: false
[2020-01-24 10:38:28] [config] lr-warmup: 0
[2020-01-24 10:38:28] [config] lr-warmup-at-reload: false
[2020-01-24 10:38:28] [config] lr-warmup-cycle: false
[2020-01-24 10:38:28] [config] lr-warmup-start-rate: 0
[2020-01-24 10:38:28] [config] max-length: 50
[2020-01-24 10:38:28] [config] max-length-crop: false
[2020-01-24 10:38:28] [config] max-length-factor: 3
[2020-01-24 10:38:28] [config] maxi-batch: 100
[2020-01-24 10:38:28] [config] maxi-batch-sort: trg
[2020-01-24 10:38:28] [config] mini-batch: 64
[2020-01-24 10:38:28] [config] mini-batch-fit: true
[2020-01-24 10:38:28] [config] mini-batch-fit-step: 10
[2020-01-24 10:38:28] [config] mini-batch-overstuff: 1
[2020-01-24 10:38:28] [config] mini-batch-track-lr: false
[2020-01-24 10:38:28] [config] mini-batch-understuff: 1
[2020-01-24 10:38:28] [config] mini-batch-warmup: 0
[2020-01-24 10:38:28] [config] mini-batch-words: 0
[2020-01-24 10:38:28] [config] mini-batch-words-ref: 0
[2020-01-24 10:38:28] [config] model: preprocess-train/model/model.npz
[2020-01-24 10:38:28] [config] multi-loss-type: sum
[2020-01-24 10:38:28] [config] multi-node: false
[2020-01-24 10:38:28] [config] multi-node-overlap: true
[2020-01-24 10:38:28] [config] n-best: false
[2020-01-24 10:38:28] [config] no-nccl: false
[2020-01-24 10:38:28] [config] no-reload: false
[2020-01-24 10:38:28] [config] no-restore-corpus: false
[2020-01-24 10:38:28] [config] normalize: 0
[2020-01-24 10:38:28] [config] normalize-gradient: false
[2020-01-24 10:38:28] [config] num-devices: 0
[2020-01-24 10:38:28] [config] optimizer: adam
[2020-01-24 10:38:28] [config] optimizer-delay: 1
[2020-01-24 10:38:28] [config] optimizer-params:
[2020-01-24 10:38:28] [config] []
[2020-01-24 10:38:28] [config] overwrite: true
[2020-01-24 10:38:28] [config] precision:
[2020-01-24 10:38:28] [config] - float32
[2020-01-24 10:38:28] [config] - float32
[2020-01-24 10:38:28] [config] - float32
[2020-01-24 10:38:28] [config] pretrained-model: ""
[2020-01-24 10:38:28] [config] quiet: false
[2020-01-24 10:38:28] [config] quiet-translation: false
[2020-01-24 10:38:28] [config] relative-paths: false
[2020-01-24 10:38:28] [config] right-left: false
[2020-01-24 10:38:28] [config] save-freq: 10000
[2020-01-24 10:38:28] [config] seed: 0
[2020-01-24 10:38:28] [config] shuffle: data
[2020-01-24 10:38:28] [config] shuffle-in-ram: false
[2020-01-24 10:38:28] [config] skip: false
[2020-01-24 10:38:28] [config] sqlite: ""
[2020-01-24 10:38:28] [config] sqlite-drop: false
[2020-01-24 10:38:28] [config] sync-sgd: false
[2020-01-24 10:38:28] [config] tempdir: /tmp
[2020-01-24 10:38:28] [config] tied-embeddings: false
[2020-01-24 10:38:28] [config] tied-embeddings-all: false
[2020-01-24 10:38:28] [config] tied-embeddings-src: false
[2020-01-24 10:38:28] [config] train-sets:
[2020-01-24 10:38:28] [config] - preprocess-train/data/europarl.tc.cs
[2020-01-24 10:38:28] [config] - preprocess-train/data/europarl.tc.en
[2020-01-24 10:38:28] [config] transformer-aan-activation: swish
[2020-01-24 10:38:28] [config] transformer-aan-depth: 2
[2020-01-24 10:38:28] [config] transformer-aan-nogate: false
[2020-01-24 10:38:28] [config] transformer-decoder-autoreg: self-attention
[2020-01-24 10:38:28] [config] transformer-depth-scaling: false
[2020-01-24 10:38:28] [config] transformer-dim-aan: 2048
[2020-01-24 10:38:28] [config] transformer-dim-ffn: 2048
[2020-01-24 10:38:28] [config] transformer-dropout: 0
[2020-01-24 10:38:28] [config] transformer-dropout-attention: 0
[2020-01-24 10:38:28] [config] transformer-dropout-ffn: 0
[2020-01-24 10:38:28] [config] transformer-ffn-activation: swish
[2020-01-24 10:38:28] [config] transformer-ffn-depth: 2
[2020-01-24 10:38:28] [config] transformer-guided-alignment-layer: last
[2020-01-24 10:38:28] [config] transformer-heads: 8
[2020-01-24 10:38:28] [config] transformer-no-projection: false
[2020-01-24 10:38:28] [config] transformer-postprocess: dan
[2020-01-24 10:38:28] [config] transformer-postprocess-emb: d
[2020-01-24 10:38:28] [config] transformer-preprocess: ""
[2020-01-24 10:38:28] [config] transformer-tied-layers:
[2020-01-24 10:38:28] [config] []
[2020-01-24 10:38:28] [config] transformer-train-position-embeddings: false
[2020-01-24 10:38:28] [config] type: s2s
[2020-01-24 10:38:28] [config] ulr: false
[2020-01-24 10:38:28] [config] ulr-dim-emb: 0
[2020-01-24 10:38:28] [config] ulr-dropout: 0
[2020-01-24 10:38:28] [config] ulr-keys-vectors: ""
[2020-01-24 10:38:28] [config] ulr-query-vectors: ""
[2020-01-24 10:38:28] [config] ulr-softmax-temperature: 1
[2020-01-24 10:38:28] [config] ulr-trainable-transformation: false
[2020-01-24 10:38:28] [config] unlikelihood-loss: false
[2020-01-24 10:38:28] [config] valid-freq: 10000
[2020-01-24 10:38:28] [config] valid-log: preprocess-train/model/valid.log
[2020-01-24 10:38:28] [config] valid-max-length: 100
[2020-01-24 10:38:28] [config] valid-metrics:
[2020-01-24 10:38:28] [config] - cross-entropy
[2020-01-24 10:38:28] [config] - translation
[2020-01-24 10:38:28] [config] valid-mini-batch: 23
[2020-01-24 10:38:28] [config] valid-script-args:
[2020-01-24 10:38:28] [config] []
[2020-01-24 10:38:28] [config] valid-script-path: bash ./preprocess-train/scripts/validate.sh
[2020-01-24 10:38:28] [config] valid-sets:
[2020-01-24 10:38:28] [config] - preprocess-train/data/valideuroparl.tc.cs
[2020-01-24 10:38:28] [config] - preprocess-train/data/valideuroparl.tc.en
[2020-01-24 10:38:28] [config] valid-translation-output: ""
[2020-01-24 10:38:28] [config] version: v1.8.33 eba7aed 2019-12-13 11:09:02 -0800
[2020-01-24 10:38:28] [config] vocabs:
[2020-01-24 10:38:28] [config] - preprocess-train/model/vocab.cs.yml
[2020-01-24 10:38:28] [config] - preprocess-train/model/vocab.en.yml
[2020-01-24 10:38:28] [config] word-penalty: 0
[2020-01-24 10:38:28] [config] workspace: 1024
[2020-01-24 10:38:28] [config] Loaded model has been created with Marian v1.8.33 eba7aed 2019-12-13 11:09:02 -0800
[2020-01-24 10:38:28] Using single-device training
[2020-01-24 10:38:28] [data] Loading vocabulary from JSON/Yaml file preprocess-train/model/vocab.cs.yml
[2020-01-24 10:38:29] [data] Setting vocabulary size for input 0 to 50000
[2020-01-24 10:38:29] [data] Loading vocabulary from JSON/Yaml file preprocess-train/model/vocab.en.yml
[2020-01-24 10:38:29] [data] Setting vocabulary size for input 1 to 50000
[2020-01-24 10:38:29] Compiled without MPI support. Falling back to FakeMPIWrapper
[2020-01-24 10:38:29] [batching] Collecting statistics for batch fitting with step size 10
[2020-01-24 10:38:30] [memory] Extending reserved space to 1024 MB (device gpu0)
[2020-01-24 10:38:31] [logits] applyLossFunction() for 1 factors
[2020-01-24 10:38:31] [memory] Reserving 422 MB, device gpu0
[2020-01-24 10:38:31] [gpu] 16-bit TensorCores enabled for float32 matrix operations
[2020-01-24 10:38:31] [memory] Reserving 422 MB, device gpu0
[2020-01-24 10:38:38] [batching] Done. Typical MB size is 1562 target words
[2020-01-24 10:38:38] [memory] Extending reserved space to 1024 MB (device gpu0)
[2020-01-24 10:38:38] Loading model from preprocess-train/model/model.npz
[2020-01-24 10:38:44] Loading Adam parameters from preprocess-train/model/model.npz.optimizer.npz
[2020-01-24 10:38:55] [memory] Reserving 844 MB, device gpu0
[2020-01-24 10:38:56] [data] Restoring the corpus state to epoch 4, batch 40000
[2020-01-24 10:38:56] [data] Shuffling data
[2020-01-24 10:38:59] [data] Done reading 640633 sentences
[2020-01-24 10:39:02] [data] Done shuffling 640633 sentences to temp files
[2020-01-24 10:39:09] Training started
[2020-01-24 10:39:09] [memory] Reserving 422 MB, device gpu0
[2020-01-24 10:39:09] [memory] Reserving 422 MB, device gpu0
[2020-01-24 10:49:00] Ep. 4 : Up. 41000 : Sen. 225,356 : Cost 36.22032547 : Time 630.31s : 1935.39 words/s
[2020-01-24 10:58:44] Ep. 4 : Up. 42000 : Sen. 274,385 : Cost 36.03122711 : Time 584.49s : 2074.00 words/s
[2020-01-24 11:08:24] Ep. 4 : Up. 43000 : Sen. 322,628 : Cost 36.29774857 : Time 580.01s : 2075.49 words/s
[2020-01-24 11:18:07] Ep. 4 : Up. 44000 : Sen. 371,371 : Cost 35.94551849 : Time 582.85s : 2074.33 words/s
[2020-01-24 11:27:50] Ep. 4 : Up. 45000 : Sen. 419,908 : Cost 36.06457520 : Time 583.19s : 2073.61 words/s
[2020-01-24 11:37:36] Ep. 4 : Up. 46000 : Sen. 468,695 : Cost 35.95717239 : Time 586.16s : 2073.60 words/s
[2020-01-24 11:47:18] Ep. 4 : Up. 47000 : Sen. 517,260 : Cost 35.63859177 : Time 581.68s : 2073.78 words/s
[2020-01-24 11:57:01] Ep. 4 : Up. 48000 : Sen. 565,605 : Cost 35.90847397 : Time 582.89s : 2070.27 words/s
[2020-01-24 12:01:55] Seen 589960 samples
[2020-01-24 12:01:55] Starting epoch 5
[2020-01-24 12:01:55] [data] Shuffling data
[2020-01-24 12:01:55] [data] Done reading 640633 sentences
[2020-01-24 12:01:58] [data] Done shuffling 640633 sentences to temp files
[2020-01-24 12:06:47] Ep. 5 : Up. 49000 : Sen. 24,115 : Cost 34.29458618 : Time 586.44s : 2056.73 words/s
[2020-01-24 12:16:34] Ep. 5 : Up. 50000 : Sen. 72,708 : Cost 32.81056976 : Time 586.83s : 2064.70 words/s
[2020-01-24 12:16:49] Saving model weights and runtime parameters to preprocess-train/model/model.npz.best-cross-entropy.npz
[2020-01-24 12:16:53] [valid] Ep. 5 : Up. 50000 : cross-entropy : 84.0588 : new best
[2020-01-24 12:16:53] Translating validation set...
[2020-01-24 12:17:05] Best translation 0 : anti @-@ Semitic re @-@ election strategy
[2020-01-24 12:17:05] Best translation 1 : the Moldovan leaders have not made their policy necessary to combat electoral fraud .
[2020-01-24 12:17:05] Best translation 2 : however , the Maltese Centre says such deception as a myth that electoral fraud in the USA is less frequent than the number of people who die after the intervention of the Taliban .
[2020-01-24 12:17:05] Best translation 3 : after all , there have been only 300 cases of electoral fraud in the US over the last ten years .
[2020-01-24 12:17:05] Best translation 4 : one is certain : these new provisions will have a negative impact on electoral participation .
[2020-01-24 12:17:05] Best translation 5 : in this sense , this measure has partly led the US democratic system .
[2020-01-24 12:17:05] Best translation 10 : the new electoral law requires voters to submit a identity card with photographs and proof of the American State citizenship .
[2020-01-24 12:17:05] Best translation 20 : in addition , it is often more than a hundred dollars to purchase the right card .
[2020-01-24 12:17:05] Best translation 40 : on the other hand , those who have imagined 76 % of all voters were only 46 % of the voters they voted in advance .
[2020-01-24 12:17:05] Best translation 80 : cigarettes cost 85 % of cancer cases .
[2020-01-24 12:17:05] Best translation 160 : ' throughout life , one has been taught that a half @-@ year corner , ' A ' Pierre Market ' , was called upon to be invited to be in favour .
[2020-01-24 12:17:15] Best translation 320 : this means that it is very efficient and that it gives a great deal of confusion .
[2020-01-24 12:17:26] Best translation 640 : the Russian Republic has presented a proposal to tighten the law in relation to the owners of civil weapons .
[2020-01-24 12:18:03] Best translation 1280 : in a nutshell , the current society is based on the unscrupulous sovereignty of their own benefit , no matter the most committed interest .
[2020-01-24 12:19:12] Best translation 2560 : after the very large sulphur blocks , which have brought together to top up to 70 percent , is to be kept on the shoulders of a two @-@ year @-@ old shop in the dark .
[2020-01-24 12:19:27] Total translation time: 154.11523s
[2020-01-24 12:19:29] [valid] Ep. 5 : Up. 50000 : translation : 15.5 : stalled 1 times (last best: 15.55)
[2020-01-24 12:19:29] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-24 12:19:33] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-24 12:29:28] Ep. 5 : Up. 51000 : Sen. 121,342 : Cost 32.70168686 : Time 773.40s : 1558.97 words/s
[2020-01-24 12:39:09] Ep. 5 : Up. 52000 : Sen. 170,373 : Cost 32.70759583 : Time 581.09s : 2086.40 words/s
[2020-01-24 12:48:51] Ep. 5 : Up. 53000 : Sen. 219,090 : Cost 32.97969055 : Time 582.13s : 2089.42 words/s
[2020-01-24 12:58:31] Ep. 5 : Up. 54000 : Sen. 267,455 : Cost 33.31706619 : Time 580.17s : 2089.61 words/s
[2020-01-24 13:08:11] Ep. 5 : Up. 55000 : Sen. 316,234 : Cost 33.03322220 : Time 580.08s : 2085.28 words/s
[2020-01-24 13:18:00] Ep. 5 : Up. 56000 : Sen. 364,953 : Cost 33.05397415 : Time 588.51s : 2066.18 words/s
[2020-01-24 13:27:46] Ep. 5 : Up. 57000 : Sen. 413,680 : Cost 33.04795074 : Time 586.54s : 2061.25 words/s
[2020-01-24 13:37:33] Ep. 5 : Up. 58000 : Sen. 462,319 : Cost 33.11449051 : Time 586.81s : 2069.39 words/s
[2020-01-24 13:47:44] Ep. 5 : Up. 59000 : Sen. 511,004 : Cost 32.94503021 : Time 611.18s : 1979.17 words/s
[2020-01-24 13:57:50] Ep. 5 : Up. 60000 : Sen. 559,590 : Cost 33.27229309 : Time 605.89s : 2003.65 words/s
[2020-01-24 13:58:05] Saving model weights and runtime parameters to preprocess-train/model/model.npz.best-cross-entropy.npz
[2020-01-24 13:58:10] [valid] Ep. 5 : Up. 60000 : cross-entropy : 82.5176 : new best
[2020-01-24 13:58:10] Translating validation set...
[2020-01-24 13:58:21] Best translation 0 : the strategy by the rebel parties against the Christmas box .
[2020-01-24 13:58:21] Best translation 1 : the leaders attach their policy to the fight against electoral fraud .
[2020-01-24 13:58:21] Best translation 2 : however , the Sixth Centre says such a deception as a myth that electoral fraud in the US is less frequent than the number of people who die after intervention .
[2020-01-24 13:58:21] Best translation 3 : after all , there have been only 300 cases of electoral fraud in the US over the last 10 years .
[2020-01-24 13:58:21] Best translation 4 : one thing is certain : these new provisions will have a negative impact on election participation .
[2020-01-24 13:58:21] Best translation 5 : in this sense , this measure was partly the American democratic system .
[2020-01-24 13:58:21] Best translation 10 : the new electoral law requires voters to submit the identity of photographs and proof of US State citizenship .
[2020-01-24 13:58:21] Best translation 20 : in addition , the purchase required identity card costs more than 100 dollars .
[2020-01-24 13:58:22] Best translation 40 : on the other hand , those who asked about 76 % of all voters were only 46 % of the electorate who voted in advance .
[2020-01-24 13:58:22] Best translation 80 : cigarettes cost 85 % of cancer cases .
[2020-01-24 13:58:22] Best translation 160 : 'in a whole of life , one teaches himself that he was a half @-@ year @-@ in @-@ way street , a shot in the devices of Paris @-@ nilly .
[2020-01-24 13:58:32] Best translation 320 : this means that it is very efficient and that he has done a great deal .
[2020-01-24 13:58:44] Best translation 640 : indeed , Russia is presenting a proposal to tighten up the law in relation to the owners of civil weapons .
[2020-01-24 13:59:24] Best translation 1280 : in writing . - ( DE ) The end of the current company is based on unscrupulous self @-@ benefit , regardless of the self @-@ interest interest .
[2020-01-24 14:00:40] Best translation 2560 : after the large @-@ scale sulphur blocks , which put together a total of 70 percent , are being held on the shoulders of a two @-@ year @-@ old @-@ old @-@ old shop .
[2020-01-24 14:00:56] Total translation time: 166.50363s
[2020-01-24 14:00:57] Saving model weights and runtime parameters to preprocess-train/model/model.npz.best-translation.npz
[2020-01-24 14:01:02] [valid] Ep. 5 : Up. 60000 : translation : 16.08 : new best
[2020-01-24 14:01:02] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-24 14:01:07] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-24 14:07:31] Seen 589960 samples
[2020-01-24 14:07:31] Starting epoch 6
[2020-01-24 14:07:31] Training finished
[2020-01-24 14:07:47] Saving model weights and runtime parameters to preprocess-train/model/model.npz.best-cross-entropy.npz
[2020-01-24 14:07:51] [valid] Ep. 6 : Up. 60621 : cross-entropy : 82.5103 : new best
[2020-01-24 14:07:51] Translating validation set...
[2020-01-24 14:08:03] Best translation 0 : strategy of the rebel parties against the Christmas box
[2020-01-24 14:08:03] Best translation 1 : the leaders attach their policy to the fight against electoral fraud .
[2020-01-24 14:08:03] Best translation 2 : however , the centre @-@ based centre is the myth that electoral fraud in the US is less frequent than the number of people who die after intervention .
[2020-01-24 14:08:03] Best translation 3 : after all , there have only been 300 cases of electoral fraud in the US over the last ten years .
[2020-01-24 14:08:03] Best translation 4 : one thing is certain : these new provisions will have a negative impact on electoral participation .
[2020-01-24 14:08:03] Best translation 5 : in this sense , this measure is partly the American democratic system .
[2020-01-24 14:08:03] Best translation 10 : new electoral laws require voters to submit the identity of photographs and proof of US citizenship .
[2020-01-24 14:08:03] Best translation 20 : in addition , the purchase required for identity cards costs more than 100 dollars .
[2020-01-24 14:08:03] Best translation 40 : on the other hand , those who voted for 76 % of all voters were only 46 % of the electorate who voted in advance .
[2020-01-24 14:08:03] Best translation 80 : cigarettes cost 85 % of cancer cases .
[2020-01-24 14:08:03] Best translation 160 : 'We learn that , in the whole of life , you learn the fact that the six @-@ year @-@ old white seat , ' On Pierre Morel , is on the part of the S & D facility .
[2020-01-24 14:08:13] Best translation 320 : this means that it is highly efficient and that it gives a great deal of confusion .
[2020-01-24 14:08:26] Best translation 640 : Russia is presenting a proposal to tighten up the law with regard to the owners of civil weapons .
[2020-01-24 14:09:06] Best translation 1280 : in writing . - ( DE ) First of all , it is the current society that is based on unscrupulous self @-@ esteem , regardless of the opposing interest .
[2020-01-24 14:10:23] Best translation 2560 : after the very large sulphur blocks , which put together a total of 70 percent , are being held on the shoulders of a two @-@ year @-@ old @-@ old @-@ old shop .
[2020-01-24 14:10:41] Total translation time: 169.37866s
[2020-01-24 14:10:42] [valid] Ep. 6 : Up. 60621 : translation : 15.87 : stalled 1 times (last best: 16.08)
[2020-01-24 14:10:42] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-24 14:10:46] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-24 14:47:50] [marian] Marian v1.8.33 eba7aed 2019-12-13 11:09:02 -0800
[2020-01-24 14:47:50] [marian] Running on jakub-Lenovo-ideapad-330-15ICH as process 9079 with command line:
[2020-01-24 14:47:50] [marian] /home/jakub/TAU/TAU_24/marian-dev/build/marian --devices 0 --type s2s --early-stopping 10 --model preprocess-train/model/model.npz --train-sets preprocess-train/data/europarl.tc.cs preprocess-train/data/europarl.tc.en --vocabs preprocess-train/model/vocab.cs.yml preprocess-train/model/vocab.en.yml --dim-vocabs 50000 50000 --mini-batch-fit -w 1024 --valid-freq 10000 --save-freq 10000 --disp-freq 1000 --valid-mini-batch 23 --valid-max-length 100 --valid-metrics cross-entropy translation --valid-sets preprocess-train/data/valideuroparl.tc.cs preprocess-train/data/valideuroparl.tc.en --valid-script-path 'bash ./preprocess-train/scripts/validate.sh' --log preprocess-train/model/train.log --valid-log preprocess-train/model/valid.log --overwrite --keep-best
[2020-01-24 14:47:50] [config] after-batches: 0
[2020-01-24 14:47:50] [config] after-epochs: 5
[2020-01-24 14:47:50] [config] all-caps-every: 0
[2020-01-24 14:47:50] [config] allow-unk: false
[2020-01-24 14:47:50] [config] authors: false
[2020-01-24 14:47:50] [config] beam-size: 12
[2020-01-24 14:47:50] [config] bert-class-symbol: "[CLS]"
[2020-01-24 14:47:50] [config] bert-mask-symbol: "[MASK]"
[2020-01-24 14:47:50] [config] bert-masking-fraction: 0.15
[2020-01-24 14:47:50] [config] bert-sep-symbol: "[SEP]"
[2020-01-24 14:47:50] [config] bert-train-type-embeddings: true
[2020-01-24 14:47:50] [config] bert-type-vocab-size: 2
[2020-01-24 14:47:50] [config] cite: false
[2020-01-24 14:47:50] [config] clip-gemm: 0
[2020-01-24 14:47:50] [config] clip-norm: 1
[2020-01-24 14:47:50] [config] cost-scaling:
[2020-01-24 14:47:50] [config] []
[2020-01-24 14:47:50] [config] cost-type: ce-mean
[2020-01-24 14:47:50] [config] cpu-threads: 0
[2020-01-24 14:47:50] [config] data-weighting: ""
[2020-01-24 14:47:50] [config] data-weighting-type: sentence
[2020-01-24 14:47:50] [config] dec-cell: gru
[2020-01-24 14:47:50] [config] dec-cell-base-depth: 2
[2020-01-24 14:47:50] [config] dec-cell-high-depth: 1
[2020-01-24 14:47:50] [config] dec-depth: 1
[2020-01-24 14:47:50] [config] devices:
[2020-01-24 14:47:50] [config] - 0
[2020-01-24 14:47:50] [config] dim-emb: 512
[2020-01-24 14:47:50] [config] dim-rnn: 1024
[2020-01-24 14:47:50] [config] dim-vocabs:
[2020-01-24 14:47:50] [config] - 50000
[2020-01-24 14:47:50] [config] - 50000
[2020-01-24 14:47:50] [config] disp-first: 0
[2020-01-24 14:47:50] [config] disp-freq: 1000
[2020-01-24 14:47:50] [config] disp-label-counts: false
[2020-01-24 14:47:50] [config] dropout-rnn: 0
[2020-01-24 14:47:50] [config] dropout-src: 0
[2020-01-24 14:47:50] [config] dropout-trg: 0
[2020-01-24 14:47:50] [config] dump-config: ""
[2020-01-24 14:47:50] [config] early-stopping: 10
[2020-01-24 14:47:50] [config] embedding-fix-src: false
[2020-01-24 14:47:50] [config] embedding-fix-trg: false
[2020-01-24 14:47:50] [config] embedding-normalization: false
[2020-01-24 14:47:50] [config] embedding-vectors:
[2020-01-24 14:47:50] [config] []
[2020-01-24 14:47:50] [config] enc-cell: gru
[2020-01-24 14:47:50] [config] enc-cell-depth: 1
[2020-01-24 14:47:50] [config] enc-depth: 1
[2020-01-24 14:47:50] [config] enc-type: bidirectional
[2020-01-24 14:47:50] [config] english-title-case-every: 0
[2020-01-24 14:47:50] [config] exponential-smoothing: 0
[2020-01-24 14:47:50] [config] factor-weight: 1
[2020-01-24 14:47:50] [config] grad-dropping-momentum: 0
[2020-01-24 14:47:50] [config] grad-dropping-rate: 0
[2020-01-24 14:47:50] [config] grad-dropping-warmup: 100
[2020-01-24 14:47:50] [config] gradient-checkpointing: false
[2020-01-24 14:47:50] [config] guided-alignment: none
[2020-01-24 14:47:50] [config] guided-alignment-cost: mse
[2020-01-24 14:47:50] [config] guided-alignment-weight: 0.1
[2020-01-24 14:47:50] [config] ignore-model-config: false
[2020-01-24 14:47:50] [config] input-types:
[2020-01-24 14:47:50] [config] []
[2020-01-24 14:47:50] [config] interpolate-env-vars: false
[2020-01-24 14:47:50] [config] keep-best: true
[2020-01-24 14:47:50] [config] label-smoothing: 0
[2020-01-24 14:47:50] [config] layer-normalization: false
[2020-01-24 14:47:50] [config] learn-rate: 0.0001
[2020-01-24 14:47:50] [config] lemma-dim-emb: 0
[2020-01-24 14:47:50] [config] log: preprocess-train/model/train.log
[2020-01-24 14:47:50] [config] log-level: info
[2020-01-24 14:47:50] [config] log-time-zone: ""
[2020-01-24 14:47:50] [config] lr-decay: 0
[2020-01-24 14:47:50] [config] lr-decay-freq: 50000
[2020-01-24 14:47:50] [config] lr-decay-inv-sqrt:
[2020-01-24 14:47:50] [config] - 0
[2020-01-24 14:47:50] [config] lr-decay-repeat-warmup: false
[2020-01-24 14:47:50] [config] lr-decay-reset-optimizer: false
[2020-01-24 14:47:50] [config] lr-decay-start:
[2020-01-24 14:47:50] [config] - 10
[2020-01-24 14:47:50] [config] - 1
[2020-01-24 14:47:50] [config] lr-decay-strategy: epoch+stalled
[2020-01-24 14:47:50] [config] lr-report: false
[2020-01-24 14:47:50] [config] lr-warmup: 0
[2020-01-24 14:47:50] [config] lr-warmup-at-reload: false
[2020-01-24 14:47:50] [config] lr-warmup-cycle: false
[2020-01-24 14:47:50] [config] lr-warmup-start-rate: 0
[2020-01-24 14:47:50] [config] max-length: 50
[2020-01-24 14:47:50] [config] max-length-crop: false
[2020-01-24 14:47:50] [config] max-length-factor: 3
[2020-01-24 14:47:50] [config] maxi-batch: 100
[2020-01-24 14:47:50] [config] maxi-batch-sort: trg
[2020-01-24 14:47:50] [config] mini-batch: 64
[2020-01-24 14:47:50] [config] mini-batch-fit: true
[2020-01-24 14:47:50] [config] mini-batch-fit-step: 10
[2020-01-24 14:47:50] [config] mini-batch-overstuff: 1
[2020-01-24 14:47:50] [config] mini-batch-track-lr: false
[2020-01-24 14:47:50] [config] mini-batch-understuff: 1
[2020-01-24 14:47:50] [config] mini-batch-warmup: 0
[2020-01-24 14:47:50] [config] mini-batch-words: 0
[2020-01-24 14:47:50] [config] mini-batch-words-ref: 0
[2020-01-24 14:47:50] [config] model: preprocess-train/model/model.npz
[2020-01-24 14:47:50] [config] multi-loss-type: sum
[2020-01-24 14:47:50] [config] multi-node: false
[2020-01-24 14:47:50] [config] multi-node-overlap: true
[2020-01-24 14:47:50] [config] n-best: false
[2020-01-24 14:47:50] [config] no-nccl: false
[2020-01-24 14:47:50] [config] no-reload: false
[2020-01-24 14:47:50] [config] no-restore-corpus: false
[2020-01-24 14:47:50] [config] normalize: 0
[2020-01-24 14:47:50] [config] normalize-gradient: false
[2020-01-24 14:47:50] [config] num-devices: 0
[2020-01-24 14:47:50] [config] optimizer: adam
[2020-01-24 14:47:50] [config] optimizer-delay: 1
[2020-01-24 14:47:50] [config] optimizer-params:
[2020-01-24 14:47:50] [config] []
[2020-01-24 14:47:50] [config] overwrite: true
[2020-01-24 14:47:50] [config] precision:
[2020-01-24 14:47:50] [config] - float32
[2020-01-24 14:47:50] [config] - float32
[2020-01-24 14:47:50] [config] - float32
[2020-01-24 14:47:50] [config] pretrained-model: ""
[2020-01-24 14:47:50] [config] quiet: false
[2020-01-24 14:47:50] [config] quiet-translation: false
[2020-01-24 14:47:50] [config] relative-paths: false
[2020-01-24 14:47:50] [config] right-left: false
[2020-01-24 14:47:50] [config] save-freq: 10000
[2020-01-24 14:47:50] [config] seed: 0
[2020-01-24 14:47:50] [config] shuffle: data
[2020-01-24 14:47:50] [config] shuffle-in-ram: false
[2020-01-24 14:47:50] [config] skip: false
[2020-01-24 14:47:50] [config] sqlite: ""
[2020-01-24 14:47:50] [config] sqlite-drop: false
[2020-01-24 14:47:50] [config] sync-sgd: false
[2020-01-24 14:47:50] [config] tempdir: /tmp
[2020-01-24 14:47:50] [config] tied-embeddings: false
[2020-01-24 14:47:50] [config] tied-embeddings-all: false
[2020-01-24 14:47:50] [config] tied-embeddings-src: false
[2020-01-24 14:47:50] [config] train-sets:
[2020-01-24 14:47:50] [config] - preprocess-train/data/europarl.tc.cs
[2020-01-24 14:47:50] [config] - preprocess-train/data/europarl.tc.en
[2020-01-24 14:47:50] [config] transformer-aan-activation: swish
[2020-01-24 14:47:50] [config] transformer-aan-depth: 2
[2020-01-24 14:47:50] [config] transformer-aan-nogate: false
[2020-01-24 14:47:50] [config] transformer-decoder-autoreg: self-attention
[2020-01-24 14:47:50] [config] transformer-depth-scaling: false
[2020-01-24 14:47:50] [config] transformer-dim-aan: 2048
[2020-01-24 14:47:50] [config] transformer-dim-ffn: 2048
[2020-01-24 14:47:50] [config] transformer-dropout: 0
[2020-01-24 14:47:50] [config] transformer-dropout-attention: 0
[2020-01-24 14:47:50] [config] transformer-dropout-ffn: 0
[2020-01-24 14:47:50] [config] transformer-ffn-activation: swish
[2020-01-24 14:47:50] [config] transformer-ffn-depth: 2
[2020-01-24 14:47:50] [config] transformer-guided-alignment-layer: last
[2020-01-24 14:47:50] [config] transformer-heads: 8
[2020-01-24 14:47:50] [config] transformer-no-projection: false
[2020-01-24 14:47:50] [config] transformer-postprocess: dan
[2020-01-24 14:47:50] [config] transformer-postprocess-emb: d
[2020-01-24 14:47:50] [config] transformer-preprocess: ""
[2020-01-24 14:47:50] [config] transformer-tied-layers:
[2020-01-24 14:47:50] [config] []
[2020-01-24 14:47:50] [config] transformer-train-position-embeddings: false
[2020-01-24 14:47:50] [config] type: s2s
[2020-01-24 14:47:50] [config] ulr: false
[2020-01-24 14:47:50] [config] ulr-dim-emb: 0
[2020-01-24 14:47:50] [config] ulr-dropout: 0
[2020-01-24 14:47:50] [config] ulr-keys-vectors: ""
[2020-01-24 14:47:50] [config] ulr-query-vectors: ""
[2020-01-24 14:47:50] [config] ulr-softmax-temperature: 1
[2020-01-24 14:47:50] [config] ulr-trainable-transformation: false
[2020-01-24 14:47:50] [config] unlikelihood-loss: false
[2020-01-24 14:47:50] [config] valid-freq: 10000
[2020-01-24 14:47:50] [config] valid-log: preprocess-train/model/valid.log
[2020-01-24 14:47:50] [config] valid-max-length: 100
[2020-01-24 14:47:50] [config] valid-metrics:
[2020-01-24 14:47:50] [config] - cross-entropy
[2020-01-24 14:47:50] [config] - translation
[2020-01-24 14:47:50] [config] valid-mini-batch: 23
[2020-01-24 14:47:50] [config] valid-script-args:
[2020-01-24 14:47:50] [config] []
[2020-01-24 14:47:50] [config] valid-script-path: bash ./preprocess-train/scripts/validate.sh
[2020-01-24 14:47:50] [config] valid-sets:
[2020-01-24 14:47:50] [config] - preprocess-train/data/valideuroparl.tc.cs
[2020-01-24 14:47:50] [config] - preprocess-train/data/valideuroparl.tc.en
[2020-01-24 14:47:50] [config] valid-translation-output: ""
[2020-01-24 14:47:50] [config] version: v1.8.33 eba7aed 2019-12-13 11:09:02 -0800
[2020-01-24 14:47:50] [config] vocabs:
[2020-01-24 14:47:50] [config] - preprocess-train/model/vocab.cs.yml
[2020-01-24 14:47:50] [config] - preprocess-train/model/vocab.en.yml
[2020-01-24 14:47:50] [config] word-penalty: 0
[2020-01-24 14:47:50] [config] workspace: 1024
[2020-01-24 14:47:50] [config] Loaded model has been created with Marian v1.8.33 eba7aed 2019-12-13 11:09:02 -0800
[2020-01-24 14:47:50] Using single-device training
[2020-01-24 14:47:50] [data] Loading vocabulary from JSON/Yaml file preprocess-train/model/vocab.cs.yml
[2020-01-24 14:47:50] [data] Setting vocabulary size for input 0 to 50000
[2020-01-24 14:47:50] [data] Loading vocabulary from JSON/Yaml file preprocess-train/model/vocab.en.yml
[2020-01-24 14:47:51] [data] Setting vocabulary size for input 1 to 50000
[2020-01-24 14:47:51] Compiled without MPI support. Falling back to FakeMPIWrapper
[2020-01-24 14:47:51] [batching] Collecting statistics for batch fitting with step size 10
[2020-01-24 14:47:52] [memory] Extending reserved space to 1024 MB (device gpu0)
[2020-01-24 14:47:52] [logits] applyLossFunction() for 1 factors
[2020-01-24 14:47:52] [memory] Reserving 422 MB, device gpu0
[2020-01-24 14:47:52] [gpu] 16-bit TensorCores enabled for float32 matrix operations
[2020-01-24 14:47:52] [memory] Reserving 422 MB, device gpu0
[2020-01-24 14:48:00] [batching] Done. Typical MB size is 1562 target words
[2020-01-24 14:48:00] [memory] Extending reserved space to 1024 MB (device gpu0)
[2020-01-24 14:48:00] Loading model from preprocess-train/model/model.npz
[2020-01-24 14:48:01] Loading Adam parameters from preprocess-train/model/model.npz.optimizer.npz
[2020-01-24 14:48:14] [memory] Reserving 844 MB, device gpu0
[2020-01-24 14:48:14] [data] Restoring the corpus state to epoch 6, batch 60621
[2020-01-24 14:48:14] [data] Shuffling data
[2020-01-24 14:48:15] [data] Done reading 640633 sentences
[2020-01-24 14:48:18] [data] Done shuffling 640633 sentences to temp files
[2020-01-24 14:48:18] Training started
[2020-01-24 14:48:18] Training finished
[2020-01-24 14:53:03] [marian] Marian v1.8.33 eba7aed 2019-12-13 11:09:02 -0800
[2020-01-24 14:53:03] [marian] Running on jakub-Lenovo-ideapad-330-15ICH as process 9261 with command line:
[2020-01-24 14:53:03] [marian] /home/jakub/TAU/TAU_24/marian-dev/build/marian --devices 0 --type s2s --early-stopping 10 --model preprocess-train/model/model.npz --train-sets preprocess-train/data/europarl.tc.cs preprocess-train/data/europarl.tc.en --vocabs preprocess-train/model/vocab.cs.yml preprocess-train/model/vocab.en.yml --dim-vocabs 50000 50000 --mini-batch-fit -w 1024 --valid-freq 10000 --save-freq 10000 --disp-freq 1000 --valid-mini-batch 23 --valid-max-length 100 --valid-metrics cross-entropy translation --valid-sets preprocess-train/data/valideuroparl.tc.cs preprocess-train/data/valideuroparl.tc.en --valid-script-path 'bash ./preprocess-train/scripts/validate.sh' --log preprocess-train/model/train.log --valid-log preprocess-train/model/valid.log --overwrite --keep-best
[2020-01-24 14:53:03] [config] after-batches: 0
[2020-01-24 14:53:03] [config] after-epochs: 5
[2020-01-24 14:53:03] [config] all-caps-every: 0
[2020-01-24 14:53:03] [config] allow-unk: false
[2020-01-24 14:53:03] [config] authors: false
[2020-01-24 14:53:03] [config] beam-size: 12
[2020-01-24 14:53:03] [config] bert-class-symbol: "[CLS]"
[2020-01-24 14:53:03] [config] bert-mask-symbol: "[MASK]"
[2020-01-24 14:53:03] [config] bert-masking-fraction: 0.15
[2020-01-24 14:53:03] [config] bert-sep-symbol: "[SEP]"
[2020-01-24 14:53:03] [config] bert-train-type-embeddings: true
[2020-01-24 14:53:03] [config] bert-type-vocab-size: 2
[2020-01-24 14:53:03] [config] cite: false
[2020-01-24 14:53:03] [config] clip-gemm: 0
[2020-01-24 14:53:03] [config] clip-norm: 1
[2020-01-24 14:53:03] [config] cost-scaling:
[2020-01-24 14:53:03] [config] []
[2020-01-24 14:53:03] [config] cost-type: ce-mean
[2020-01-24 14:53:03] [config] cpu-threads: 0
[2020-01-24 14:53:03] [config] data-weighting: ""
[2020-01-24 14:53:03] [config] data-weighting-type: sentence
[2020-01-24 14:53:03] [config] dec-cell: gru
[2020-01-24 14:53:03] [config] dec-cell-base-depth: 2
[2020-01-24 14:53:03] [config] dec-cell-high-depth: 1
[2020-01-24 14:53:03] [config] dec-depth: 1
[2020-01-24 14:53:03] [config] devices:
[2020-01-24 14:53:03] [config] - 0
[2020-01-24 14:53:03] [config] dim-emb: 512
[2020-01-24 14:53:03] [config] dim-rnn: 1024
[2020-01-24 14:53:03] [config] dim-vocabs:
[2020-01-24 14:53:03] [config] - 50000
[2020-01-24 14:53:03] [config] - 50000
[2020-01-24 14:53:03] [config] disp-first: 0
[2020-01-24 14:53:03] [config] disp-freq: 1000
[2020-01-24 14:53:03] [config] disp-label-counts: false
[2020-01-24 14:53:03] [config] dropout-rnn: 0
[2020-01-24 14:53:03] [config] dropout-src: 0
[2020-01-24 14:53:03] [config] dropout-trg: 0
[2020-01-24 14:53:03] [config] dump-config: ""
[2020-01-24 14:53:03] [config] early-stopping: 10
[2020-01-24 14:53:03] [config] embedding-fix-src: false
[2020-01-24 14:53:03] [config] embedding-fix-trg: false
[2020-01-24 14:53:03] [config] embedding-normalization: false
[2020-01-24 14:53:03] [config] embedding-vectors:
[2020-01-24 14:53:03] [config] []
[2020-01-24 14:53:03] [config] enc-cell: gru
[2020-01-24 14:53:03] [config] enc-cell-depth: 1
[2020-01-24 14:53:03] [config] enc-depth: 1
[2020-01-24 14:53:03] [config] enc-type: bidirectional
[2020-01-24 14:53:03] [config] english-title-case-every: 0
[2020-01-24 14:53:03] [config] exponential-smoothing: 0
[2020-01-24 14:53:03] [config] factor-weight: 1
[2020-01-24 14:53:03] [config] grad-dropping-momentum: 0
[2020-01-24 14:53:03] [config] grad-dropping-rate: 0
[2020-01-24 14:53:03] [config] grad-dropping-warmup: 100
[2020-01-24 14:53:03] [config] gradient-checkpointing: false
[2020-01-24 14:53:03] [config] guided-alignment: none
[2020-01-24 14:53:03] [config] guided-alignment-cost: mse
[2020-01-24 14:53:03] [config] guided-alignment-weight: 0.1
[2020-01-24 14:53:03] [config] ignore-model-config: false
[2020-01-24 14:53:03] [config] input-types:
[2020-01-24 14:53:03] [config] []
[2020-01-24 14:53:03] [config] interpolate-env-vars: false
[2020-01-24 14:53:03] [config] keep-best: true
[2020-01-24 14:53:03] [config] label-smoothing: 0
[2020-01-24 14:53:03] [config] layer-normalization: false
[2020-01-24 14:53:03] [config] learn-rate: 0.0001
[2020-01-24 14:53:03] [config] lemma-dim-emb: 0
[2020-01-24 14:53:03] [config] log: preprocess-train/model/train.log
[2020-01-24 14:53:03] [config] log-level: info
[2020-01-24 14:53:03] [config] log-time-zone: ""
[2020-01-24 14:53:03] [config] lr-decay: 0
[2020-01-24 14:53:03] [config] lr-decay-freq: 50000
[2020-01-24 14:53:03] [config] lr-decay-inv-sqrt:
[2020-01-24 14:53:03] [config] - 0
[2020-01-24 14:53:03] [config] lr-decay-repeat-warmup: false
[2020-01-24 14:53:03] [config] lr-decay-reset-optimizer: false
[2020-01-24 14:53:03] [config] lr-decay-start:
[2020-01-24 14:53:03] [config] - 10
[2020-01-24 14:53:03] [config] - 1
[2020-01-24 14:53:03] [config] lr-decay-strategy: epoch+stalled
[2020-01-24 14:53:03] [config] lr-report: false
[2020-01-24 14:53:03] [config] lr-warmup: 0
[2020-01-24 14:53:03] [config] lr-warmup-at-reload: false
[2020-01-24 14:53:03] [config] lr-warmup-cycle: false
[2020-01-24 14:53:03] [config] lr-warmup-start-rate: 0
[2020-01-24 14:53:03] [config] max-length: 50
[2020-01-24 14:53:03] [config] max-length-crop: false
[2020-01-24 14:53:03] [config] max-length-factor: 3
[2020-01-24 14:53:03] [config] maxi-batch: 100
[2020-01-24 14:53:03] [config] maxi-batch-sort: trg
[2020-01-24 14:53:03] [config] mini-batch: 64
[2020-01-24 14:53:03] [config] mini-batch-fit: true
[2020-01-24 14:53:03] [config] mini-batch-fit-step: 10
[2020-01-24 14:53:03] [config] mini-batch-overstuff: 1
[2020-01-24 14:53:03] [config] mini-batch-track-lr: false
[2020-01-24 14:53:03] [config] mini-batch-understuff: 1
[2020-01-24 14:53:03] [config] mini-batch-warmup: 0
[2020-01-24 14:53:03] [config] mini-batch-words: 0
[2020-01-24 14:53:03] [config] mini-batch-words-ref: 0
[2020-01-24 14:53:03] [config] model: preprocess-train/model/model.npz
[2020-01-24 14:53:03] [config] multi-loss-type: sum
[2020-01-24 14:53:03] [config] multi-node: false
[2020-01-24 14:53:03] [config] multi-node-overlap: true
[2020-01-24 14:53:03] [config] n-best: false
[2020-01-24 14:53:03] [config] no-nccl: false
[2020-01-24 14:53:03] [config] no-reload: false
[2020-01-24 14:53:03] [config] no-restore-corpus: false
[2020-01-24 14:53:03] [config] normalize: 0
[2020-01-24 14:53:03] [config] normalize-gradient: false
[2020-01-24 14:53:03] [config] num-devices: 0
[2020-01-24 14:53:03] [config] optimizer: adam
[2020-01-24 14:53:03] [config] optimizer-delay: 1
[2020-01-24 14:53:03] [config] optimizer-params:
[2020-01-24 14:53:03] [config] []
[2020-01-24 14:53:03] [config] overwrite: true
[2020-01-24 14:53:03] [config] precision:
[2020-01-24 14:53:03] [config] - float32
[2020-01-24 14:53:03] [config] - float32
[2020-01-24 14:53:03] [config] - float32
[2020-01-24 14:53:03] [config] pretrained-model: ""
[2020-01-24 14:53:03] [config] quiet: false
[2020-01-24 14:53:03] [config] quiet-translation: false
[2020-01-24 14:53:03] [config] relative-paths: false
[2020-01-24 14:53:03] [config] right-left: false
[2020-01-24 14:53:03] [config] save-freq: 10000
[2020-01-24 14:53:03] [config] seed: 0
[2020-01-24 14:53:03] [config] shuffle: data
[2020-01-24 14:53:03] [config] shuffle-in-ram: false
[2020-01-24 14:53:03] [config] skip: false
[2020-01-24 14:53:03] [config] sqlite: ""
[2020-01-24 14:53:03] [config] sqlite-drop: false
[2020-01-24 14:53:03] [config] sync-sgd: false
[2020-01-24 14:53:03] [config] tempdir: /tmp
[2020-01-24 14:53:03] [config] tied-embeddings: false
[2020-01-24 14:53:03] [config] tied-embeddings-all: false
[2020-01-24 14:53:03] [config] tied-embeddings-src: false
[2020-01-24 14:53:03] [config] train-sets:
[2020-01-24 14:53:03] [config] - preprocess-train/data/europarl.tc.cs
[2020-01-24 14:53:03] [config] - preprocess-train/data/europarl.tc.en
[2020-01-24 14:53:03] [config] transformer-aan-activation: swish
[2020-01-24 14:53:03] [config] transformer-aan-depth: 2
[2020-01-24 14:53:03] [config] transformer-aan-nogate: false
[2020-01-24 14:53:03] [config] transformer-decoder-autoreg: self-attention
[2020-01-24 14:53:03] [config] transformer-depth-scaling: false
[2020-01-24 14:53:03] [config] transformer-dim-aan: 2048
[2020-01-24 14:53:03] [config] transformer-dim-ffn: 2048
[2020-01-24 14:53:03] [config] transformer-dropout: 0
[2020-01-24 14:53:03] [config] transformer-dropout-attention: 0
[2020-01-24 14:53:03] [config] transformer-dropout-ffn: 0
[2020-01-24 14:53:03] [config] transformer-ffn-activation: swish
[2020-01-24 14:53:03] [config] transformer-ffn-depth: 2
[2020-01-24 14:53:03] [config] transformer-guided-alignment-layer: last
[2020-01-24 14:53:03] [config] transformer-heads: 8
[2020-01-24 14:53:03] [config] transformer-no-projection: false
[2020-01-24 14:53:03] [config] transformer-postprocess: dan
[2020-01-24 14:53:03] [config] transformer-postprocess-emb: d
[2020-01-24 14:53:03] [config] transformer-preprocess: ""
[2020-01-24 14:53:03] [config] transformer-tied-layers:
[2020-01-24 14:53:03] [config] []
[2020-01-24 14:53:03] [config] transformer-train-position-embeddings: false
[2020-01-24 14:53:03] [config] type: s2s
[2020-01-24 14:53:03] [config] ulr: false
[2020-01-24 14:53:03] [config] ulr-dim-emb: 0
[2020-01-24 14:53:03] [config] ulr-dropout: 0
[2020-01-24 14:53:03] [config] ulr-keys-vectors: ""
[2020-01-24 14:53:03] [config] ulr-query-vectors: ""
[2020-01-24 14:53:03] [config] ulr-softmax-temperature: 1
[2020-01-24 14:53:03] [config] ulr-trainable-transformation: false
[2020-01-24 14:53:03] [config] unlikelihood-loss: false
[2020-01-24 14:53:03] [config] valid-freq: 10000
[2020-01-24 14:53:03] [config] valid-log: preprocess-train/model/valid.log
[2020-01-24 14:53:03] [config] valid-max-length: 100
[2020-01-24 14:53:03] [config] valid-metrics:
[2020-01-24 14:53:03] [config] - cross-entropy
[2020-01-24 14:53:03] [config] - translation
[2020-01-24 14:53:03] [config] valid-mini-batch: 23
[2020-01-24 14:53:03] [config] valid-script-args:
[2020-01-24 14:53:03] [config] []
[2020-01-24 14:53:03] [config] valid-script-path: bash ./preprocess-train/scripts/validate.sh
[2020-01-24 14:53:03] [config] valid-sets:
[2020-01-24 14:53:03] [config] - preprocess-train/data/valideuroparl.tc.cs
[2020-01-24 14:53:03] [config] - preprocess-train/data/valideuroparl.tc.en
[2020-01-24 14:53:03] [config] valid-translation-output: ""
[2020-01-24 14:53:03] [config] version: v1.8.33 eba7aed 2019-12-13 11:09:02 -0800
[2020-01-24 14:53:03] [config] vocabs:
[2020-01-24 14:53:03] [config] - preprocess-train/model/vocab.cs.yml
[2020-01-24 14:53:03] [config] - preprocess-train/model/vocab.en.yml
[2020-01-24 14:53:03] [config] word-penalty: 0
[2020-01-24 14:53:03] [config] workspace: 1024
[2020-01-24 14:53:03] [config] Loaded model has been created with Marian v1.8.33 eba7aed 2019-12-13 11:09:02 -0800
[2020-01-24 14:53:03] Using single-device training
[2020-01-24 14:53:03] [data] Loading vocabulary from JSON/Yaml file preprocess-train/model/vocab.cs.yml
[2020-01-24 14:53:03] [data] Setting vocabulary size for input 0 to 50000
[2020-01-24 14:53:03] [data] Loading vocabulary from JSON/Yaml file preprocess-train/model/vocab.en.yml
[2020-01-24 14:53:04] [data] Setting vocabulary size for input 1 to 50000
[2020-01-24 14:53:04] Compiled without MPI support. Falling back to FakeMPIWrapper
[2020-01-24 14:53:04] [batching] Collecting statistics for batch fitting with step size 10
[2020-01-24 14:53:05] [memory] Extending reserved space to 1024 MB (device gpu0)
[2020-01-24 14:53:05] [logits] applyLossFunction() for 1 factors
[2020-01-24 14:53:05] [memory] Reserving 422 MB, device gpu0
[2020-01-24 14:53:05] [gpu] 16-bit TensorCores enabled for float32 matrix operations
[2020-01-24 14:53:05] [memory] Reserving 422 MB, device gpu0
[2020-01-24 14:53:12] [batching] Done. Typical MB size is 1562 target words
[2020-01-24 14:53:13] [memory] Extending reserved space to 1024 MB (device gpu0)
[2020-01-24 14:53:13] Loading model from preprocess-train/model/model.npz
[2020-01-24 14:53:14] Loading Adam parameters from preprocess-train/model/model.npz.optimizer.npz
[2020-01-24 14:53:16] [memory] Reserving 844 MB, device gpu0
[2020-01-24 14:53:17] [data] Restoring the corpus state to epoch 6, batch 60621
[2020-01-24 14:53:17] [data] Shuffling data
[2020-01-24 14:53:17] [data] Done reading 640633 sentences
[2020-01-24 14:53:20] [data] Done shuffling 640633 sentences to temp files
[2020-01-24 14:53:20] Training started
[2020-01-24 14:53:20] Training finished
[2020-01-24 14:54:40] [marian] Marian v1.8.33 eba7aed 2019-12-13 11:09:02 -0800
[2020-01-24 14:54:40] [marian] Running on jakub-Lenovo-ideapad-330-15ICH as process 9316 with command line:
[2020-01-24 14:54:40] [marian] /home/jakub/TAU/TAU_24/marian-dev/build/marian --devices 0 --type s2s --early-stopping 10 -e 15 --model preprocess-train/model/model.npz --train-sets preprocess-train/data/europarl.tc.cs preprocess-train/data/europarl.tc.en --vocabs preprocess-train/model/vocab.cs.yml preprocess-train/model/vocab.en.yml --dim-vocabs 50000 50000 --mini-batch-fit -w 1024 --valid-freq 10000 --save-freq 10000 --disp-freq 1000 --valid-mini-batch 23 --valid-max-length 100 --valid-metrics cross-entropy translation --valid-sets preprocess-train/data/valideuroparl.tc.cs preprocess-train/data/valideuroparl.tc.en --valid-script-path 'bash ./preprocess-train/scripts/validate.sh' --log preprocess-train/model/train.log --valid-log preprocess-train/model/valid.log --overwrite --keep-best
[2020-01-24 14:54:40] [config] after-batches: 0
[2020-01-24 14:54:40] [config] after-epochs: 15
[2020-01-24 14:54:40] [config] all-caps-every: 0
[2020-01-24 14:54:40] [config] allow-unk: false
[2020-01-24 14:54:40] [config] authors: false
[2020-01-24 14:54:40] [config] beam-size: 12
[2020-01-24 14:54:40] [config] bert-class-symbol: "[CLS]"
[2020-01-24 14:54:40] [config] bert-mask-symbol: "[MASK]"
[2020-01-24 14:54:40] [config] bert-masking-fraction: 0.15
[2020-01-24 14:54:40] [config] bert-sep-symbol: "[SEP]"
[2020-01-24 14:54:40] [config] bert-train-type-embeddings: true
[2020-01-24 14:54:40] [config] bert-type-vocab-size: 2
[2020-01-24 14:54:40] [config] cite: false
[2020-01-24 14:54:40] [config] clip-gemm: 0
[2020-01-24 14:54:40] [config] clip-norm: 1
[2020-01-24 14:54:40] [config] cost-scaling:
[2020-01-24 14:54:40] [config] []
[2020-01-24 14:54:40] [config] cost-type: ce-mean
[2020-01-24 14:54:40] [config] cpu-threads: 0
[2020-01-24 14:54:40] [config] data-weighting: ""
[2020-01-24 14:54:40] [config] data-weighting-type: sentence
[2020-01-24 14:54:40] [config] dec-cell: gru
[2020-01-24 14:54:40] [config] dec-cell-base-depth: 2
[2020-01-24 14:54:40] [config] dec-cell-high-depth: 1
[2020-01-24 14:54:40] [config] dec-depth: 1
[2020-01-24 14:54:40] [config] devices:
[2020-01-24 14:54:40] [config] - 0
[2020-01-24 14:54:40] [config] dim-emb: 512
[2020-01-24 14:54:40] [config] dim-rnn: 1024
[2020-01-24 14:54:40] [config] dim-vocabs:
[2020-01-24 14:54:40] [config] - 50000
[2020-01-24 14:54:40] [config] - 50000
[2020-01-24 14:54:40] [config] disp-first: 0
[2020-01-24 14:54:40] [config] disp-freq: 1000
[2020-01-24 14:54:40] [config] disp-label-counts: false
[2020-01-24 14:54:40] [config] dropout-rnn: 0
[2020-01-24 14:54:40] [config] dropout-src: 0
[2020-01-24 14:54:40] [config] dropout-trg: 0
[2020-01-24 14:54:40] [config] dump-config: ""
[2020-01-24 14:54:40] [config] early-stopping: 10
[2020-01-24 14:54:40] [config] embedding-fix-src: false
[2020-01-24 14:54:40] [config] embedding-fix-trg: false
[2020-01-24 14:54:40] [config] embedding-normalization: false
[2020-01-24 14:54:40] [config] embedding-vectors:
[2020-01-24 14:54:40] [config] []
[2020-01-24 14:54:40] [config] enc-cell: gru
[2020-01-24 14:54:40] [config] enc-cell-depth: 1
[2020-01-24 14:54:40] [config] enc-depth: 1
[2020-01-24 14:54:40] [config] enc-type: bidirectional
[2020-01-24 14:54:40] [config] english-title-case-every: 0
[2020-01-24 14:54:40] [config] exponential-smoothing: 0
[2020-01-24 14:54:40] [config] factor-weight: 1
[2020-01-24 14:54:40] [config] grad-dropping-momentum: 0
[2020-01-24 14:54:40] [config] grad-dropping-rate: 0
[2020-01-24 14:54:40] [config] grad-dropping-warmup: 100
[2020-01-24 14:54:40] [config] gradient-checkpointing: false
[2020-01-24 14:54:40] [config] guided-alignment: none
[2020-01-24 14:54:40] [config] guided-alignment-cost: mse
[2020-01-24 14:54:40] [config] guided-alignment-weight: 0.1
[2020-01-24 14:54:40] [config] ignore-model-config: false
[2020-01-24 14:54:40] [config] input-types:
[2020-01-24 14:54:40] [config] []
[2020-01-24 14:54:40] [config] interpolate-env-vars: false
[2020-01-24 14:54:40] [config] keep-best: true
[2020-01-24 14:54:40] [config] label-smoothing: 0
[2020-01-24 14:54:40] [config] layer-normalization: false
[2020-01-24 14:54:40] [config] learn-rate: 0.0001
[2020-01-24 14:54:40] [config] lemma-dim-emb: 0
[2020-01-24 14:54:40] [config] log: preprocess-train/model/train.log
[2020-01-24 14:54:40] [config] log-level: info
[2020-01-24 14:54:40] [config] log-time-zone: ""
[2020-01-24 14:54:40] [config] lr-decay: 0
[2020-01-24 14:54:40] [config] lr-decay-freq: 50000
[2020-01-24 14:54:40] [config] lr-decay-inv-sqrt:
[2020-01-24 14:54:40] [config] - 0
[2020-01-24 14:54:40] [config] lr-decay-repeat-warmup: false
[2020-01-24 14:54:40] [config] lr-decay-reset-optimizer: false
[2020-01-24 14:54:40] [config] lr-decay-start:
[2020-01-24 14:54:40] [config] - 10
[2020-01-24 14:54:40] [config] - 1
[2020-01-24 14:54:40] [config] lr-decay-strategy: epoch+stalled
[2020-01-24 14:54:40] [config] lr-report: false
[2020-01-24 14:54:40] [config] lr-warmup: 0
[2020-01-24 14:54:40] [config] lr-warmup-at-reload: false
[2020-01-24 14:54:40] [config] lr-warmup-cycle: false
[2020-01-24 14:54:40] [config] lr-warmup-start-rate: 0
[2020-01-24 14:54:40] [config] max-length: 50
[2020-01-24 14:54:40] [config] max-length-crop: false
[2020-01-24 14:54:40] [config] max-length-factor: 3
[2020-01-24 14:54:40] [config] maxi-batch: 100
[2020-01-24 14:54:40] [config] maxi-batch-sort: trg
[2020-01-24 14:54:40] [config] mini-batch: 64
[2020-01-24 14:54:40] [config] mini-batch-fit: true
[2020-01-24 14:54:40] [config] mini-batch-fit-step: 10
[2020-01-24 14:54:40] [config] mini-batch-overstuff: 1
[2020-01-24 14:54:40] [config] mini-batch-track-lr: false
[2020-01-24 14:54:40] [config] mini-batch-understuff: 1
[2020-01-24 14:54:40] [config] mini-batch-warmup: 0
[2020-01-24 14:54:40] [config] mini-batch-words: 0
[2020-01-24 14:54:40] [config] mini-batch-words-ref: 0
[2020-01-24 14:54:40] [config] model: preprocess-train/model/model.npz
[2020-01-24 14:54:40] [config] multi-loss-type: sum
[2020-01-24 14:54:40] [config] multi-node: false
[2020-01-24 14:54:40] [config] multi-node-overlap: true
[2020-01-24 14:54:40] [config] n-best: false
[2020-01-24 14:54:40] [config] no-nccl: false
[2020-01-24 14:54:40] [config] no-reload: false
[2020-01-24 14:54:40] [config] no-restore-corpus: false
[2020-01-24 14:54:40] [config] normalize: 0
[2020-01-24 14:54:40] [config] normalize-gradient: false
[2020-01-24 14:54:40] [config] num-devices: 0
[2020-01-24 14:54:40] [config] optimizer: adam
[2020-01-24 14:54:40] [config] optimizer-delay: 1
[2020-01-24 14:54:40] [config] optimizer-params:
[2020-01-24 14:54:40] [config] []
[2020-01-24 14:54:40] [config] overwrite: true
[2020-01-24 14:54:40] [config] precision:
[2020-01-24 14:54:40] [config] - float32
[2020-01-24 14:54:40] [config] - float32
[2020-01-24 14:54:40] [config] - float32
[2020-01-24 14:54:40] [config] pretrained-model: ""
[2020-01-24 14:54:40] [config] quiet: false
[2020-01-24 14:54:40] [config] quiet-translation: false
[2020-01-24 14:54:40] [config] relative-paths: false
[2020-01-24 14:54:40] [config] right-left: false
[2020-01-24 14:54:40] [config] save-freq: 10000
[2020-01-24 14:54:40] [config] seed: 0
[2020-01-24 14:54:40] [config] shuffle: data
[2020-01-24 14:54:40] [config] shuffle-in-ram: false
[2020-01-24 14:54:40] [config] skip: false
[2020-01-24 14:54:40] [config] sqlite: ""
[2020-01-24 14:54:40] [config] sqlite-drop: false
[2020-01-24 14:54:40] [config] sync-sgd: false
[2020-01-24 14:54:40] [config] tempdir: /tmp
[2020-01-24 14:54:40] [config] tied-embeddings: false
[2020-01-24 14:54:40] [config] tied-embeddings-all: false
[2020-01-24 14:54:40] [config] tied-embeddings-src: false
[2020-01-24 14:54:40] [config] train-sets:
[2020-01-24 14:54:40] [config] - preprocess-train/data/europarl.tc.cs
[2020-01-24 14:54:40] [config] - preprocess-train/data/europarl.tc.en
[2020-01-24 14:54:40] [config] transformer-aan-activation: swish
[2020-01-24 14:54:40] [config] transformer-aan-depth: 2
[2020-01-24 14:54:40] [config] transformer-aan-nogate: false
[2020-01-24 14:54:40] [config] transformer-decoder-autoreg: self-attention
[2020-01-24 14:54:40] [config] transformer-depth-scaling: false
[2020-01-24 14:54:40] [config] transformer-dim-aan: 2048
[2020-01-24 14:54:40] [config] transformer-dim-ffn: 2048
[2020-01-24 14:54:40] [config] transformer-dropout: 0
[2020-01-24 14:54:40] [config] transformer-dropout-attention: 0
[2020-01-24 14:54:40] [config] transformer-dropout-ffn: 0
[2020-01-24 14:54:40] [config] transformer-ffn-activation: swish
[2020-01-24 14:54:40] [config] transformer-ffn-depth: 2
[2020-01-24 14:54:40] [config] transformer-guided-alignment-layer: last
[2020-01-24 14:54:40] [config] transformer-heads: 8
[2020-01-24 14:54:40] [config] transformer-no-projection: false
[2020-01-24 14:54:40] [config] transformer-postprocess: dan
[2020-01-24 14:54:40] [config] transformer-postprocess-emb: d
[2020-01-24 14:54:40] [config] transformer-preprocess: ""
[2020-01-24 14:54:40] [config] transformer-tied-layers:
[2020-01-24 14:54:40] [config] []
[2020-01-24 14:54:40] [config] transformer-train-position-embeddings: false
[2020-01-24 14:54:40] [config] type: s2s
[2020-01-24 14:54:40] [config] ulr: false
[2020-01-24 14:54:40] [config] ulr-dim-emb: 0
[2020-01-24 14:54:40] [config] ulr-dropout: 0
[2020-01-24 14:54:40] [config] ulr-keys-vectors: ""
[2020-01-24 14:54:40] [config] ulr-query-vectors: ""
[2020-01-24 14:54:40] [config] ulr-softmax-temperature: 1
[2020-01-24 14:54:40] [config] ulr-trainable-transformation: false
[2020-01-24 14:54:40] [config] unlikelihood-loss: false
[2020-01-24 14:54:40] [config] valid-freq: 10000
[2020-01-24 14:54:40] [config] valid-log: preprocess-train/model/valid.log
[2020-01-24 14:54:40] [config] valid-max-length: 100
[2020-01-24 14:54:40] [config] valid-metrics:
[2020-01-24 14:54:40] [config] - cross-entropy
[2020-01-24 14:54:40] [config] - translation
[2020-01-24 14:54:40] [config] valid-mini-batch: 23
[2020-01-24 14:54:40] [config] valid-script-args:
[2020-01-24 14:54:40] [config] []
[2020-01-24 14:54:40] [config] valid-script-path: bash ./preprocess-train/scripts/validate.sh
[2020-01-24 14:54:40] [config] valid-sets:
[2020-01-24 14:54:40] [config] - preprocess-train/data/valideuroparl.tc.cs
[2020-01-24 14:54:40] [config] - preprocess-train/data/valideuroparl.tc.en
[2020-01-24 14:54:40] [config] valid-translation-output: ""
[2020-01-24 14:54:40] [config] version: v1.8.33 eba7aed 2019-12-13 11:09:02 -0800
[2020-01-24 14:54:40] [config] vocabs:
[2020-01-24 14:54:40] [config] - preprocess-train/model/vocab.cs.yml
[2020-01-24 14:54:40] [config] - preprocess-train/model/vocab.en.yml
[2020-01-24 14:54:40] [config] word-penalty: 0
[2020-01-24 14:54:40] [config] workspace: 1024
[2020-01-24 14:54:40] [config] Loaded model has been created with Marian v1.8.33 eba7aed 2019-12-13 11:09:02 -0800
[2020-01-24 14:54:40] Using single-device training
[2020-01-24 14:54:40] [data] Loading vocabulary from JSON/Yaml file preprocess-train/model/vocab.cs.yml
[2020-01-24 14:54:41] [data] Setting vocabulary size for input 0 to 50000
[2020-01-24 14:54:41] [data] Loading vocabulary from JSON/Yaml file preprocess-train/model/vocab.en.yml
[2020-01-24 14:54:41] [data] Setting vocabulary size for input 1 to 50000
[2020-01-24 14:54:41] Compiled without MPI support. Falling back to FakeMPIWrapper
[2020-01-24 14:54:41] [batching] Collecting statistics for batch fitting with step size 10
[2020-01-24 14:54:42] [memory] Extending reserved space to 1024 MB (device gpu0)
[2020-01-24 14:54:42] [logits] applyLossFunction() for 1 factors
[2020-01-24 14:54:42] [memory] Reserving 422 MB, device gpu0
[2020-01-24 14:54:42] [gpu] 16-bit TensorCores enabled for float32 matrix operations
[2020-01-24 14:54:42] [memory] Reserving 422 MB, device gpu0
[2020-01-24 14:54:50] [batching] Done. Typical MB size is 1562 target words
[2020-01-24 14:54:50] [memory] Extending reserved space to 1024 MB (device gpu0)
[2020-01-24 14:54:50] Loading model from preprocess-train/model/model.npz
[2020-01-24 14:54:51] Loading Adam parameters from preprocess-train/model/model.npz.optimizer.npz
[2020-01-24 14:54:53] [memory] Reserving 844 MB, device gpu0
[2020-01-24 14:54:54] [data] Restoring the corpus state to epoch 6, batch 60621
[2020-01-24 14:54:54] [data] Shuffling data
[2020-01-24 14:54:54] [data] Done reading 640633 sentences
[2020-01-24 14:54:57] [data] Done shuffling 640633 sentences to temp files
[2020-01-24 14:54:57] Training started
[2020-01-24 14:54:58] [memory] Reserving 422 MB, device gpu0
[2020-01-24 14:54:58] [memory] Reserving 422 MB, device gpu0
[2020-01-24 14:58:51] Ep. 6 : Up. 61000 : Sen. 18,537 : Cost 31.69468117 : Time 249.70s : 4851.62 words/s
[2020-01-24 15:08:28] Ep. 6 : Up. 62000 : Sen. 67,262 : Cost 29.80109596 : Time 577.50s : 2090.54 words/s
[2020-01-24 15:18:08] Ep. 6 : Up. 63000 : Sen. 115,710 : Cost 30.35162354 : Time 580.22s : 2086.23 words/s
[2020-01-24 15:27:49] Ep. 6 : Up. 64000 : Sen. 164,631 : Cost 30.20695686 : Time 580.26s : 2091.23 words/s
[2020-01-24 15:37:30] Ep. 6 : Up. 65000 : Sen. 213,304 : Cost 30.36400223 : Time 580.88s : 2086.78 words/s
[2020-01-24 15:47:06] Ep. 6 : Up. 66000 : Sen. 261,950 : Cost 30.36856270 : Time 576.86s : 2091.11 words/s
[2020-01-24 15:57:15] Ep. 6 : Up. 67000 : Sen. 310,606 : Cost 30.58700371 : Time 608.79s : 1987.38 words/s
[2020-01-24 16:07:13] Ep. 6 : Up. 68000 : Sen. 359,155 : Cost 30.64333725 : Time 597.70s : 2029.55 words/s
[2020-01-24 16:16:57] Ep. 6 : Up. 69000 : Sen. 407,954 : Cost 30.80709457 : Time 584.52s : 2079.99 words/s
[2020-01-24 16:26:38] Ep. 6 : Up. 70000 : Sen. 456,826 : Cost 30.59236717 : Time 580.83s : 2088.72 words/s
[2020-01-24 16:26:53] Saving model weights and runtime parameters to preprocess-train/model/model.npz.best-cross-entropy.npz
[2020-01-24 16:26:57] [valid] Ep. 6 : Up. 70000 : cross-entropy : 82.2896 : new best
[2020-01-24 16:26:57] Translating validation set...
[2020-01-24 16:27:08] Best translation 0 : the Anti @-@ election strategy
[2020-01-24 16:27:08] Best translation 1 : the leaders do not recognise their policy of combating electoral fraud .
[2020-01-24 16:27:08] Best translation 2 : however , the Sixth Centre claims such a scam as a myth that electoral fraud is less frequent in the USA than the number of persons dying after the intervention .
[2020-01-24 16:27:08] Best translation 3 : after all , the last ten years have recorded only 300 cases of electoral fraud in the US .
[2020-01-24 16:27:08] Best translation 4 : one thing is certain : these new provisions will have a negative impact on voter turnout .
[2020-01-24 16:27:08] Best translation 5 : in this sense , these measures were partly exacerbated by the American democratic system .
[2020-01-24 16:27:08] Best translation 10 : new electoral laws require the voters to submit the identity card with photographs and proof of US State citizenship .
[2020-01-24 16:27:08] Best translation 20 : in addition , the purchase required card card costs more than USD 100 .
[2020-01-24 16:27:09] Best translation 40 : on the other hand , those who imagined 76 % of all voters were only 46 % of the electorate in advance .
[2020-01-24 16:27:09] Best translation 80 : cigarettes are worth 85 % of cancer cases .
[2020-01-24 16:27:09] Best translation 160 : ' throughout life , you learn the fact that a half @-@ year stage , ' Don Pierre & D , is being called upon by Mrs Betancourt .
[2020-01-24 16:27:18] Best translation 320 : this means that it is very efficient , and that he has done a great deal .
[2020-01-24 16:27:29] Best translation 640 : Russian Russia is presenting a proposal to tighten up the law in relation to the owners of civil weapons .
[2020-01-24 16:28:06] Best translation 1280 : as a result , as a result , the current society is based on an unscrupulous self @-@ benefit from no matter of self @-@ interest .
[2020-01-24 16:29:16] Best translation 2560 : in the wake of the heavy blocks of sulphur , which combined up to 70 percent , the Dalai Lama is on the shoulders of a two @-@ year @-@ old @-@ old @-@ old shop .
[2020-01-24 16:29:32] Total translation time: 154.15843s
[2020-01-24 16:29:33] [valid] Ep. 6 : Up. 70000 : translation : 16.03 : stalled 2 times (last best: 16.08)
[2020-01-24 16:29:33] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-24 16:29:37] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-24 16:39:28] Ep. 6 : Up. 71000 : Sen. 505,159 : Cost 30.75662041 : Time 770.03s : 1561.69 words/s
[2020-01-24 16:49:26] Ep. 6 : Up. 72000 : Sen. 553,807 : Cost 30.66852951 : Time 597.50s : 2019.24 words/s
[2020-01-24 16:56:55] Seen 589960 samples
[2020-01-24 16:56:55] Starting epoch 7
[2020-01-24 16:56:55] [data] Shuffling data
[2020-01-24 16:56:56] [data] Done reading 640633 sentences
[2020-01-24 16:56:59] [data] Done shuffling 640633 sentences to temp files
[2020-01-24 16:59:27] Ep. 7 : Up. 73000 : Sen. 11,990 : Cost 30.13134384 : Time 601.05s : 2006.13 words/s
[2020-01-24 17:09:27] Ep. 7 : Up. 74000 : Sen. 60,596 : Cost 27.64407921 : Time 599.80s : 2015.66 words/s
[2020-01-24 17:19:28] Ep. 7 : Up. 75000 : Sen. 109,372 : Cost 27.89711571 : Time 601.10s : 2018.30 words/s
[2020-01-24 17:29:32] Ep. 7 : Up. 76000 : Sen. 158,224 : Cost 28.00063705 : Time 604.34s : 2006.95 words/s
[2020-01-24 17:39:40] Ep. 7 : Up. 77000 : Sen. 207,172 : Cost 28.07950783 : Time 608.28s : 1999.00 words/s
[2020-01-24 17:49:42] Ep. 7 : Up. 78000 : Sen. 255,811 : Cost 28.33088112 : Time 601.15s : 2015.02 words/s
[2020-01-24 17:59:41] Ep. 7 : Up. 79000 : Sen. 304,401 : Cost 28.32571793 : Time 599.08s : 2017.40 words/s
[2020-01-24 18:09:34] Ep. 7 : Up. 80000 : Sen. 353,174 : Cost 28.44856453 : Time 593.35s : 2031.97 words/s
[2020-01-24 18:09:49] [valid] Ep. 7 : Up. 80000 : cross-entropy : 82.905 : stalled 1 times (last best: 82.2896)
[2020-01-24 18:09:49] Translating validation set...
[2020-01-24 18:10:00] Best translation 0 : the strategy of the Republican Party against Administrative Santos
[2020-01-24 18:10:00] Best translation 1 : the leaders attach their policy to the fight against electoral fraud .
[2020-01-24 18:10:00] Best translation 2 : however , the Maltese Centre claims such a scam as a myth that electoral fraud in the US is less frequent than the number of people who die after intervention .
[2020-01-24 18:10:00] Best translation 3 : after all , only 300 cases of electoral fraud in the US have been recorded over the last ten years .
[2020-01-24 18:10:00] Best translation 4 : one thing is certain : these new provisions will have a negative impact on voter turnout .
[2020-01-24 18:10:00] Best translation 5 : in this sense , these measures are partly the American democratic system .
[2020-01-24 18:10:00] Best translation 10 : the new electoral laws require the voters to submit the identity and photographs of the American State citizenship .
[2020-01-24 18:10:00] Best translation 20 : in addition , the purchase of the required identity card costs often over 100 dollars .
[2020-01-24 18:10:00] Best translation 40 : on the other hand , those who found 76 % of all voters were only 46 % of the voters they voted in advance .
[2020-01-24 18:10:00] Best translation 80 : cigarettes are responsible for 85 % of the cases of lung cancer .
[2020-01-24 18:10:00] Best translation 160 : yes , ' throughout life , it teaches us that it will not be a standing ovation , ' On Pierre Morel , in the case @-@ by @-@ case facility .
[2020-01-24 18:10:10] Best translation 320 : this means that it is very efficient and it gives a lot of clarity .
[2020-01-24 18:10:22] Best translation 640 : it is the proposal to tighten up the law in relation to the owners of civil weapons .
[2020-01-24 18:10:59] Best translation 1280 : in writing . - ( DE ) Like the current society , the conclusion is based on ruthless self @-@ benefit , irrespective of the vested interest .
[2020-01-24 18:12:10] Best translation 2560 : after the very large sulphur blocks of sulphur , which together values to 70 metres , take place at the same time as the one on the shoulders of a two @-@ year @-@ old @-@ old @-@ old @-@ old shop .
[2020-01-24 18:12:25] Total translation time: 156.06545s
[2020-01-24 18:12:26] Saving model weights and runtime parameters to preprocess-train/model/model.npz.best-translation.npz
[2020-01-24 18:12:31] [valid] Ep. 7 : Up. 80000 : translation : 16.49 : new best
[2020-01-24 18:12:31] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-24 18:12:36] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-24 18:22:48] Ep. 7 : Up. 81000 : Sen. 401,789 : Cost 28.60611916 : Time 793.82s : 1527.02 words/s
[2020-01-24 18:32:47] Ep. 7 : Up. 82000 : Sen. 450,142 : Cost 28.91402054 : Time 599.19s : 2018.74 words/s
[2020-01-24 18:42:45] Ep. 7 : Up. 83000 : Sen. 498,820 : Cost 28.54313469 : Time 598.28s : 2018.82 words/s
[2020-01-24 18:52:43] Ep. 7 : Up. 84000 : Sen. 547,327 : Cost 28.73744774 : Time 598.19s : 2021.43 words/s
[2020-01-24 19:01:31] Seen 589960 samples
[2020-01-24 19:01:31] Starting epoch 8
[2020-01-24 19:01:31] [data] Shuffling data
[2020-01-24 19:01:32] [data] Done reading 640633 sentences
[2020-01-24 19:01:34] [data] Done shuffling 640633 sentences to temp files
[2020-01-24 19:02:50] Ep. 8 : Up. 85000 : Sen. 6,041 : Cost 28.54472542 : Time 606.24s : 2003.00 words/s
[2020-01-24 19:12:51] Ep. 8 : Up. 86000 : Sen. 54,854 : Cost 25.41687965 : Time 601.54s : 2011.49 words/s
[2020-01-24 19:22:52] Ep. 8 : Up. 87000 : Sen. 103,368 : Cost 26.00359726 : Time 600.81s : 2021.25 words/s
[2020-01-24 19:32:53] Ep. 8 : Up. 88000 : Sen. 152,328 : Cost 25.94273186 : Time 601.00s : 2022.08 words/s
[2020-01-24 19:42:51] Ep. 8 : Up. 89000 : Sen. 200,803 : Cost 26.31779480 : Time 597.66s : 2019.89 words/s
[2020-01-24 19:52:50] Ep. 8 : Up. 90000 : Sen. 249,429 : Cost 26.29836464 : Time 599.09s : 2021.28 words/s
[2020-01-24 19:53:05] [valid] Ep. 8 : Up. 90000 : cross-entropy : 83.1733 : stalled 2 times (last best: 82.2896)
[2020-01-24 19:53:05] Translating validation set...
[2020-01-24 19:53:16] Best translation 0 : the strategy of the Republican Party against Shahbaz Bhatti
[2020-01-24 19:53:16] Best translation 1 : the leaders do not defend their policy on the need to combat electoral fraud .
[2020-01-24 19:53:16] Best translation 2 : however , the centre of this scam is being described as a myth that electoral fraud in the USA is less frequent than the number of people who die after the intervention .
[2020-01-24 19:53:16] Best translation 3 : after all , last ten years have recorded only 300 cases of electoral fraud in the USA .
[2020-01-24 19:53:16] Best translation 4 : one thing is certain : these new provisions will have a negative impact on voter turnout .
[2020-01-24 19:53:16] Best translation 5 : in this sense , this is partly the American democratic system .
[2020-01-24 19:53:16] Best translation 10 : the new electoral law requires the electorate to submit the identity of photographs and proof of US State citizenship .
[2020-01-24 19:53:16] Best translation 20 : in addition , the purchase of an identity card is often over a hundred dollars .
[2020-01-24 19:53:17] Best translation 40 : on the other hand , those who found 76 % % of the electorate were only 46 % of the electorate voted in advance .
[2020-01-24 19:53:17] Best translation 80 : cigarettes cost 85 % of cancer cases .
[2020-01-24 19:53:17] Best translation 160 : ' throughout life , one has been taught that a half @-@ year @-@ old dream of life , ' On Pierre Morel , to the psychologist in Erasmus .
[2020-01-24 19:53:27] Best translation 320 : this means that the co @-@ legislators are very powerful and that it gives a great deal of restraint .
[2020-01-24 19:53:38] Best translation 640 : Northern Russia is presenting a proposal to tighten up the law in relation to the owners of civil weapons .
[2020-01-24 19:54:17] Best translation 1280 : in a nutshell , the current company is based on the unscrupulous vested interests , regardless of the vested interest .
[2020-01-24 19:55:30] Best translation 2560 : after the great blocks of sulphur , which have a combined merit of up to 70 percent , were pushed on the shoulders of a two @-@ year @-@ old @-@ old @-@ old @-@ old @-@ old shop .
[2020-01-24 19:55:46] Total translation time: 161.59131s
[2020-01-24 19:55:48] [valid] Ep. 8 : Up. 90000 : translation : 16.18 : stalled 1 times (last best: 16.49)
[2020-01-24 19:55:48] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-24 19:55:52] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-24 20:06:06] Ep. 8 : Up. 91000 : Sen. 298,476 : Cost 26.42875290 : Time 796.09s : 1528.46 words/s
[2020-01-24 20:16:04] Ep. 8 : Up. 92000 : Sen. 346,831 : Cost 26.81515694 : Time 598.52s : 2018.31 words/s
[2020-01-24 20:26:11] Ep. 8 : Up. 93000 : Sen. 395,411 : Cost 26.73735619 : Time 606.42s : 1997.21 words/s
[2020-01-24 20:36:18] Ep. 8 : Up. 94000 : Sen. 444,333 : Cost 26.64360619 : Time 607.41s : 1993.21 words/s
[2020-01-24 20:46:25] Ep. 8 : Up. 95000 : Sen. 492,800 : Cost 26.89325905 : Time 607.08s : 1989.24 words/s
[2020-01-24 20:56:34] Ep. 8 : Up. 96000 : Sen. 541,127 : Cost 27.13678169 : Time 608.37s : 1988.77 words/s
[2020-01-24 21:06:41] Ep. 8 : Up. 97000 : Sen. 589,960 : Cost 26.82535172 : Time 607.43s : 1990.14 words/s
[2020-01-24 21:06:41] Seen 589960 samples
[2020-01-24 21:06:41] Starting epoch 9
[2020-01-24 21:06:41] [data] Shuffling data
[2020-01-24 21:06:41] [data] Done reading 640633 sentences
[2020-01-24 21:06:44] [data] Done shuffling 640633 sentences to temp files
[2020-01-24 21:16:52] Ep. 9 : Up. 98000 : Sen. 48,724 : Cost 23.53754044 : Time 610.85s : 1978.78 words/s
[2020-01-24 21:26:58] Ep. 9 : Up. 99000 : Sen. 96,693 : Cost 24.25844574 : Time 606.17s : 1985.99 words/s
[2020-01-24 21:36:41] Ep. 9 : Up. 100000 : Sen. 145,333 : Cost 24.07562637 : Time 582.85s : 2066.87 words/s
[2020-01-24 21:36:56] [valid] Ep. 9 : Up. 100000 : cross-entropy : 84.6166 : stalled 3 times (last best: 82.2896)
[2020-01-24 21:36:56] Translating validation set...
[2020-01-24 21:37:06] Best translation 0 : the strategy of the Republican Party against Friedrich @-@ General
[2020-01-24 21:37:06] Best translation 1 : the leaders attach their policy to the fight against electoral fraud .
[2020-01-24 21:37:06] Best translation 2 : however , the Maltese Centre claims such a scam as a myth that electoral fraud in the USA is less frequent than the number of people dying after the intervention .
[2020-01-24 21:37:06] Best translation 3 : last year , last 10 years , we have seen only 300 cases of electoral fraud in the US .
[2020-01-24 21:37:06] Best translation 4 : one thing is certain : these new provisions will have a negative impact on voter turnout .
[2020-01-24 21:37:06] Best translation 5 : in this sense , this is partly the US democratic system .
[2020-01-24 21:37:06] Best translation 10 : new electoral laws require voters to submit the identity of photographs and proof of US State citizenship .
[2020-01-24 21:37:06] Best translation 20 : in addition , the purchase requirement required is often over a hundred dollars .
[2020-01-24 21:37:06] Best translation 40 : on the other hand , there were only 46 % of the voters who voted in advance on the other side of those who found 76 % to all voters .
[2020-01-24 21:37:06] Best translation 80 : cigarettes cost 85 % of cancer cases .
[2020-01-24 21:37:06] Best translation 160 : ' throughout life , man has been learning a standing ovation , ' Kinga Pierre Morel , the psychologist Directorate @-@ oriented .
[2020-01-24 21:37:16] Best translation 320 : this means that the screen is very efficient and that it gives a lot of praise .
[2020-01-24 21:37:26] Best translation 640 : the Central European Union is proposing a tightening of the law in relation to the owners of civil weapons .
[2020-01-24 21:38:01] Best translation 1280 : as in Alsace , the current company is based on ruthless self @-@ benefit , irrespective of the vested interest .
[2020-01-24 21:39:08] Best translation 2560 : after the heavy blocks of sulphur , which are shared by up to 70 percent , are being imposed on the back of the head on the shoulders of the Suez Canal .
[2020-01-24 21:39:22] Total translation time: 146.22686s
[2020-01-24 21:39:23] [valid] Ep. 9 : Up. 100000 : translation : 16.42 : stalled 2 times (last best: 16.49)
[2020-01-24 21:39:23] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-24 21:39:27] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-24 21:49:13] Ep. 9 : Up. 101000 : Sen. 193,827 : Cost 24.28462219 : Time 752.11s : 1599.45 words/s
[2020-01-24 21:58:53] Ep. 9 : Up. 102000 : Sen. 242,122 : Cost 24.76545525 : Time 579.68s : 2080.99 words/s
[2020-01-24 22:08:34] Ep. 9 : Up. 103000 : Sen. 290,825 : Cost 24.62094307 : Time 580.78s : 2080.89 words/s
[2020-01-24 22:18:17] Ep. 9 : Up. 104000 : Sen. 339,629 : Cost 24.78282356 : Time 583.67s : 2078.21 words/s
[2020-01-24 22:27:59] Ep. 9 : Up. 105000 : Sen. 388,209 : Cost 24.91489601 : Time 581.76s : 2086.06 words/s
[2020-01-24 22:37:38] Ep. 9 : Up. 106000 : Sen. 437,036 : Cost 25.00333786 : Time 579.29s : 2089.35 words/s
[2020-01-24 22:47:22] Ep. 9 : Up. 107000 : Sen. 485,710 : Cost 25.31485939 : Time 583.74s : 2089.10 words/s
[2020-01-24 22:57:04] Ep. 9 : Up. 108000 : Sen. 534,486 : Cost 25.03606796 : Time 582.40s : 2083.26 words/s
[2020-01-24 23:06:45] Ep. 9 : Up. 109000 : Sen. 583,532 : Cost 25.16001892 : Time 580.99s : 2087.59 words/s
[2020-01-24 23:08:04] Seen 589960 samples
[2020-01-24 23:08:04] Starting epoch 10
[2020-01-24 23:08:04] [data] Shuffling data
[2020-01-24 23:08:04] [data] Done reading 640633 sentences
[2020-01-24 23:08:07] [data] Done shuffling 640633 sentences to temp files
[2020-01-24 23:16:28] Ep. 10 : Up. 110000 : Sen. 41,931 : Cost 22.45999718 : Time 582.28s : 2074.87 words/s
[2020-01-24 23:16:42] [valid] Ep. 10 : Up. 110000 : cross-entropy : 85.5421 : stalled 4 times (last best: 82.2896)
[2020-01-24 23:16:42] Translating validation set...
[2020-01-24 23:16:53] Best translation 0 : the strategy of the Republican Party against bilingual re @-@ election
[2020-01-24 23:16:53] Best translation 1 : the leaders have shown their policy necessary to combat electoral fraud .
[2020-01-24 23:16:53] Best translation 2 : however , the Maltese Centre claims such a sham with the myth that electoral fraud in the USA is less frequent than the number of people dying after the intervention .
[2020-01-24 23:16:53] Best translation 3 : last year in the last ten years , only 300 cases have been recorded in US electoral fraud .
[2020-01-24 23:16:53] Best translation 4 : one thing is certain : these new provisions will have a negative impact on voter turnout .
[2020-01-24 23:16:53] Best translation 5 : in this sense , these measures were partly regulated by the American democratic system .
[2020-01-24 23:16:53] Best translation 10 : new electoral laws require voters to submit the identity of photographs and evidence of US State citizenship .
[2020-01-24 23:16:53] Best translation 20 : furthermore , the purchase required card card costs often over 100 dollars .
[2020-01-24 23:16:54] Best translation 40 : on the other hand , those who found 76 % % of the electorate were only 46 % of the electorate in advance .
[2020-01-24 23:16:54] Best translation 80 : cigarettes are responsible for 85 % of cases of lung cancer .
[2020-01-24 23:16:54] Best translation 160 : 'i throughout life it teaches us that a 12 @-@ year @-@ old man 's standing ovation , ' On Pierre Morel , to the psychologist and to fork .
[2020-01-24 23:17:03] Best translation 320 : this means that the Eurogroup is very efficient and has a lot to do .
[2020-01-24 23:17:14] Best translation 640 : the Icelandic Russia is proposing a tightening of the law in relation to the owners of civil weapons .
[2020-01-24 23:17:49] Best translation 1280 : in writing . - ( DE ) A final word on the ground is that the current society is based on the unscrupulous values of self @-@ benefit , irrespective of the vested interest .
[2020-01-24 23:18:56] Best translation 2560 : after the very large sulphur reactors , which consist together to be 70 percent , have been pushed into the back on the shoulders of the 24 @-@ year @-@ old shop .
[2020-01-24 23:19:11] Total translation time: 148.33650s
[2020-01-24 23:19:12] [valid] Ep. 10 : Up. 110000 : translation : 16.23 : stalled 3 times (last best: 16.49)
[2020-01-24 23:19:12] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-24 23:19:16] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-24 23:29:09] Ep. 10 : Up. 111000 : Sen. 90,852 : Cost 22.20336151 : Time 761.27s : 1597.06 words/s
[2020-01-24 23:38:52] Ep. 10 : Up. 112000 : Sen. 139,408 : Cost 22.50568581 : Time 582.64s : 2075.37 words/s
[2020-01-24 23:48:33] Ep. 10 : Up. 113000 : Sen. 188,064 : Cost 22.60812378 : Time 581.50s : 2081.73 words/s
[2020-01-24 23:58:14] Ep. 10 : Up. 114000 : Sen. 236,800 : Cost 22.84693336 : Time 580.88s : 2085.12 words/s
[2020-01-25 00:07:56] Ep. 10 : Up. 115000 : Sen. 285,765 : Cost 22.76149750 : Time 581.97s : 2087.59 words/s
[2020-01-25 00:17:53] Ep. 10 : Up. 116000 : Sen. 334,454 : Cost 23.13924408 : Time 597.37s : 2032.66 words/s
[2020-01-25 00:28:11] Ep. 10 : Up. 117000 : Sen. 383,077 : Cost 23.29958725 : Time 618.08s : 1960.12 words/s
[2020-01-25 00:38:14] Ep. 10 : Up. 118000 : Sen. 431,657 : Cost 23.36029434 : Time 602.36s : 2008.72 words/s
[2020-01-25 00:47:54] Ep. 10 : Up. 119000 : Sen. 480,423 : Cost 23.55471230 : Time 580.71s : 2093.70 words/s
[2020-01-25 00:57:35] Ep. 10 : Up. 120000 : Sen. 529,225 : Cost 23.54610634 : Time 580.04s : 2087.89 words/s
[2020-01-25 00:57:49] [valid] Ep. 10 : Up. 120000 : cross-entropy : 86.5355 : stalled 5 times (last best: 82.2896)
[2020-01-25 00:57:49] Translating validation set...
[2020-01-25 00:58:00] Best translation 0 : the strategy of the Republican Party against His re @-@ election
[2020-01-25 00:58:00] Best translation 1 : the leaders have shown their policy necessary to combat electoral fraud .
[2020-01-25 00:58:00] Best translation 2 : however , the Tamil Centre claims such a con as a myth that electoral fraud in the USA is less frequent than the number of those who die after intervention .
[2020-01-25 00:58:00] Best translation 3 : after all , in the last 10 years there have been only 300 cases of electoral fraud in the USA .
[2020-01-25 00:58:00] Best translation 4 : one thing is certain : these new provisions will have a negative impact on voter turnout .
[2020-01-25 00:58:00] Best translation 5 : in this sense , the measure is partly the American democratic system .
[2020-01-25 00:58:00] Best translation 10 : the new electoral laws require voters to submit identity cards and proof of US citizenship .
[2020-01-25 00:58:00] Best translation 20 : in addition , the purchase required identity card costs more than a hundred dollars .
[2020-01-25 00:58:00] Best translation 40 : on the other side , there were only 46 % of the electorate who voted in advance , on the other side , who found 76 % to all voters .
[2020-01-25 00:58:00] Best translation 80 : cigarettes are responsible for 85 % of cases of lung cancer .
[2020-01-25 00:58:00] Best translation 160 : ' throughout life ' throughout life it teaches us that the birthplace of a parliament 's standing ovation , ' On Pierre Morel , the psychologist site in Erasmus .
[2020-01-25 00:58:10] Best translation 320 : this means that the co @-@ contractor is very powerful and he has done a great deal .
[2020-01-25 00:58:21] Best translation 640 : Central Russia proposes a proposal to tighten up the law in relation to the owners of civil weapons .
[2020-01-25 00:58:57] Best translation 1280 : in a nutshell , the current society is based on an unscrupulous vested self @-@ benefit , no matter what they want .
[2020-01-25 01:00:07] Best translation 2560 : then , on the back of the total sulphur sulphur survey , which combined up to 70 fuels , the Dalai Lama is on the shoulders of a two @-@ year @-@ old shop for the Suez Canal .
[2020-01-25 01:00:22] Total translation time: 153.01506s
[2020-01-25 01:00:23] [valid] Ep. 10 : Up. 120000 : translation : 15.92 : stalled 4 times (last best: 16.49)
[2020-01-25 01:00:23] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-25 01:00:27] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-25 01:10:15] Ep. 10 : Up. 121000 : Sen. 578,088 : Cost 23.53382111 : Time 760.37s : 1591.64 words/s
[2020-01-25 01:12:38] Seen 589960 samples
[2020-01-25 01:12:38] Starting epoch 11
[2020-01-25 01:12:38] [data] Shuffling data
[2020-01-25 01:12:39] [data] Done reading 640633 sentences
[2020-01-25 01:12:41] [data] Done shuffling 640633 sentences to temp files
[2020-01-25 01:19:53] Ep. 11 : Up. 122000 : Sen. 36,530 : Cost 21.10663223 : Time 578.30s : 2077.57 words/s
[2020-01-25 01:29:29] Ep. 11 : Up. 123000 : Sen. 84,858 : Cost 20.58423233 : Time 575.86s : 2089.45 words/s
[2020-01-25 01:39:06] Ep. 11 : Up. 124000 : Sen. 133,444 : Cost 20.71440887 : Time 576.80s : 2093.89 words/s
[2020-01-25 01:48:46] Ep. 11 : Up. 125000 : Sen. 182,632 : Cost 20.83464622 : Time 579.66s : 2096.31 words/s
[2020-01-25 01:58:23] Ep. 11 : Up. 126000 : Sen. 230,824 : Cost 21.33970642 : Time 577.83s : 2090.34 words/s
[2020-01-25 02:08:01] Ep. 11 : Up. 127000 : Sen. 279,481 : Cost 21.33987045 : Time 578.01s : 2091.42 words/s
[2020-01-25 02:17:40] Ep. 11 : Up. 128000 : Sen. 328,247 : Cost 21.47406960 : Time 578.51s : 2094.41 words/s
[2020-01-25 02:27:20] Ep. 11 : Up. 129000 : Sen. 376,849 : Cost 21.77974701 : Time 580.47s : 2094.11 words/s
[2020-01-25 02:36:56] Ep. 11 : Up. 130000 : Sen. 425,390 : Cost 21.69615936 : Time 575.66s : 2095.01 words/s
[2020-01-25 02:37:11] [valid] Ep. 11 : Up. 130000 : cross-entropy : 88.0175 : stalled 6 times (last best: 82.2896)
[2020-01-25 02:37:11] Translating validation set...
[2020-01-25 02:37:21] Best translation 0 : the strategy of the Republican Party against Vocational resumed
[2020-01-25 02:37:21] Best translation 1 : the leaders have established their policy necessary to combat electoral fraud .
[2020-01-25 02:37:21] Best translation 2 : however , the Centre of Malta refers to such a scam as a myth that electoral fraud in the USA is less frequent than that of those who die after the intervention .
[2020-01-25 02:37:21] Best translation 3 : after all , over the last 10 years there have been only 300 cases of electoral fraud in the US .
[2020-01-25 02:37:21] Best translation 4 : one thing is certain : these new provisions will have a negative impact on voter turnout .
[2020-01-25 02:37:21] Best translation 5 : in this sense , this was partly the American democratic system .
[2020-01-25 02:37:21] Best translation 10 : new electoral laws require voters to submit identity of photographs and proof of American State citizenship .
[2020-01-25 02:37:21] Best translation 20 : in addition , the purchase required identity card costs more than a hundred dollars .
[2020-01-25 02:37:21] Best translation 40 : on the other hand , I think only those who voted for 76 % of the voters were only 46 % of the voters voted in advance .
[2020-01-25 02:37:21] Best translation 80 : cigarettes cost 85 % of the cases of lung cancer .
[2020-01-25 02:37:21] Best translation 160 : ' throughout life , you learn that they were standing by a standing ovation , ' Think Pierre Morel , the questioner @-@ centred .
[2020-01-25 02:37:30] Best translation 320 : this means that the screen is very efficient and that it sets a lot of fun .
[2020-01-25 02:37:41] Best translation 640 : Northern Russia is proposing a proposal to tighten up the law with regard to the owners of civil weapons .
[2020-01-25 02:38:15] Best translation 1280 : in a nutshell , the current society is based on the unscrupulous vested interests , regardless of the vested interest .
[2020-01-25 02:39:21] Best translation 2560 : once the large sulphur blocks of the sulphur , which bring together to 70 kilo , are being held up on the back of the next two @-@ year @-@ old anti @-@ Europeans .
[2020-01-25 02:39:35] Total translation time: 144.41860s
[2020-01-25 02:39:36] [valid] Ep. 11 : Up. 130000 : translation : 15.98 : stalled 5 times (last best: 16.49)
[2020-01-25 02:39:36] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-25 02:39:41] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-25 02:49:30] Ep. 11 : Up. 131000 : Sen. 473,771 : Cost 22.03436279 : Time 753.77s : 1606.54 words/s
[2020-01-25 02:59:09] Ep. 11 : Up. 132000 : Sen. 522,827 : Cost 21.92679214 : Time 579.54s : 2102.18 words/s
[2020-01-25 03:08:46] Ep. 11 : Up. 133000 : Sen. 571,578 : Cost 21.91812897 : Time 576.38s : 2100.70 words/s
[2020-01-25 03:12:27] Seen 589960 samples
[2020-01-25 03:12:27] Starting epoch 12
[2020-01-25 03:12:27] [data] Shuffling data
[2020-01-25 03:12:27] [data] Done reading 640633 sentences
[2020-01-25 03:12:30] [data] Done shuffling 640633 sentences to temp files
[2020-01-25 03:18:30] Ep. 12 : Up. 134000 : Sen. 30,254 : Cost 20.18366241 : Time 584.68s : 2081.95 words/s
[2020-01-25 03:28:06] Ep. 12 : Up. 135000 : Sen. 79,191 : Cost 18.76207733 : Time 575.72s : 2104.40 words/s
[2020-01-25 03:37:46] Ep. 12 : Up. 136000 : Sen. 128,126 : Cost 19.09635925 : Time 579.46s : 2098.78 words/s
[2020-01-25 03:47:26] Ep. 12 : Up. 137000 : Sen. 176,725 : Cost 19.49136925 : Time 580.09s : 2097.19 words/s
[2020-01-25 03:57:02] Ep. 12 : Up. 138000 : Sen. 225,265 : Cost 19.77016830 : Time 576.58s : 2099.04 words/s
[2020-01-25 04:06:41] Ep. 12 : Up. 139000 : Sen. 274,091 : Cost 19.69299698 : Time 578.46s : 2100.04 words/s
[2020-01-25 04:16:17] Ep. 12 : Up. 140000 : Sen. 322,921 : Cost 19.80997849 : Time 576.54s : 2095.43 words/s
[2020-01-25 04:16:32] [valid] Ep. 12 : Up. 140000 : cross-entropy : 89.7736 : stalled 7 times (last best: 82.2896)
[2020-01-25 04:16:32] Translating validation set...
[2020-01-25 04:16:42] Best translation 0 : the strategy of the Republican Party against Administrative resumed
[2020-01-25 04:16:42] Best translation 1 : the most conservative leaders owe their policy necessary to combat electoral fraud .
[2020-01-25 04:16:42] Best translation 2 : however , the rebel centre referred to such a scam as a myth that electoral fraud in the USA has been less frequent than the number of people dying after the intervention .
[2020-01-25 04:16:42] Best translation 3 : after all , the last ten months have seen only 300 cases of electoral fraud in the US .
[2020-01-25 04:16:42] Best translation 4 : one thing is certain : these new provisions will have a negative impact on the turnout .
[2020-01-25 04:16:42] Best translation 5 : in this sense , this measure partially brought up the American democratic system .
[2020-01-25 04:16:42] Best translation 10 : new electoral laws require voters to submit identity of photographs and indications of American citizenship .
[2020-01-25 04:16:42] Best translation 20 : in addition , the acquisition required claim is often more than a hundred dollars .
[2020-01-25 04:16:42] Best translation 40 : on the other hand , there were only 46 % of the voters who voted in advance , on the other side , who were to present 76 % of the voters .
[2020-01-25 04:16:42] Best translation 80 : cigarettes are responsible for 85 % of cases of lung cancer .
[2020-01-25 04:16:42] Best translation 160 : employed throughout life , people learn to be the leader of an anti @-@ European man who , in his words , was talking about Pierre Werner , and the invitation to inspire them .
[2020-01-25 04:16:52] Best translation 320 : this means that the Eurogroup is very efficient and has much to do .
[2020-01-25 04:17:02] Best translation 640 : the proposal for Russia is a proposal to tighten up the law with regard to the owners of civil weapons .
[2020-01-25 04:17:37] Best translation 1280 : in writing . - ( DE ) A final word of all is that the current society is based on the unscrupulous vested interests , regardless of the interest shown .
[2020-01-25 04:18:43] Best translation 2560 : having devastated the large sulphur reactors which bring together to EUR 70 fuels , the Dalai Lama is on the shoulders of the next two @-@ year @-@ old baby with the Suez Canal .
[2020-01-25 04:18:57] Total translation time: 145.21051s
[2020-01-25 04:18:58] [valid] Ep. 12 : Up. 140000 : translation : 15.72 : stalled 6 times (last best: 16.49)
[2020-01-25 04:18:58] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-25 04:19:02] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-25 04:28:53] Ep. 12 : Up. 141000 : Sen. 371,503 : Cost 20.17894363 : Time 755.51s : 1603.71 words/s
[2020-01-25 04:38:31] Ep. 12 : Up. 142000 : Sen. 420,250 : Cost 20.14449501 : Time 578.28s : 2090.64 words/s
[2020-01-25 04:48:13] Ep. 12 : Up. 143000 : Sen. 468,918 : Cost 20.53544807 : Time 582.12s : 2091.63 words/s
[2020-01-25 04:57:51] Ep. 12 : Up. 144000 : Sen. 517,638 : Cost 20.40444565 : Time 578.21s : 2093.51 words/s
[2020-01-25 05:07:29] Ep. 12 : Up. 145000 : Sen. 566,302 : Cost 20.58670425 : Time 577.35s : 2094.33 words/s
[2020-01-25 05:12:11] Seen 589960 samples
[2020-01-25 05:12:11] Starting epoch 13
[2020-01-25 05:12:11] [data] Shuffling data
[2020-01-25 05:12:11] [data] Done reading 640633 sentences
[2020-01-25 05:12:14] [data] Done shuffling 640633 sentences to temp files
[2020-01-25 05:17:10] Ep. 13 : Up. 146000 : Sen. 24,869 : Cost 18.86958504 : Time 581.38s : 2077.97 words/s
[2020-01-25 05:26:47] Ep. 13 : Up. 147000 : Sen. 73,589 : Cost 17.46876335 : Time 577.07s : 2094.03 words/s
[2020-01-25 05:36:24] Ep. 13 : Up. 148000 : Sen. 122,149 : Cost 17.61222839 : Time 576.81s : 2094.92 words/s
[2020-01-25 05:46:04] Ep. 13 : Up. 149000 : Sen. 171,203 : Cost 17.81441879 : Time 579.77s : 2092.45 words/s
[2020-01-25 05:55:45] Ep. 13 : Up. 150000 : Sen. 220,053 : Cost 18.15541649 : Time 581.22s : 2091.22 words/s
[2020-01-25 05:55:59] [valid] Ep. 13 : Up. 150000 : cross-entropy : 91.4267 : stalled 8 times (last best: 82.2896)
[2020-01-25 05:55:59] Translating validation set...
[2020-01-25 05:56:10] Best translation 0 : the strategy of the Republican Party against International resumed
[2020-01-25 05:56:10] Best translation 1 : the leaders will see their policy necessary to combat electoral fraud .
[2020-01-25 05:56:10] Best translation 2 : however , the rebel centre referred to such a deception as a myth that electoral fraud in the USA is less frequent than the number of those who die after the intervention of the Székely people .
[2020-01-25 05:56:10] Best translation 3 : in fact , over the last 10 years , there have only been 300 cases of electoral fraud in the US .
[2020-01-25 05:56:10] Best translation 4 : one thing is certain : these new provisions will have a negative effect on voter turnout .
[2020-01-25 05:56:10] Best translation 5 : in this sense , this is partly a simple American democratic system .
[2020-01-25 05:56:10] Best translation 10 : new electoral laws require the voters to present the identity of photographs and proof of US state citizenship .
[2020-01-25 05:56:10] Best translation 20 : in addition , the acquisition required claim is often more than a hundred dollars .
[2020-01-25 05:56:10] Best translation 40 : on the other hand , there were only 46 % of the electorate who voted in advance for the arrival of 76 % of all voters .
[2020-01-25 05:56:10] Best translation 80 : cigarettes cost 85 % of the cases of lung cancer .
[2020-01-25 05:56:10] Best translation 160 : " throughout life , one has been learning a standing ovation , ' On Pierre Morel , and the passion in Erasmus facilities .
[2020-01-25 05:56:19] Best translation 320 : this means that the Eurogroup is very efficient and has a lot to do .
[2020-01-25 05:56:30] Best translation 640 : Central Russia is tabling a proposal to tighten up the law with regard to the owners of civil weapons .
[2020-01-25 05:57:06] Best translation 1280 : in writing . - ( DE ) Like the current society , it is also based on an unscrupulous appetite for self @-@ benefit .
[2020-01-25 05:58:12] Best translation 2560 : after the heavy blocks of sulphur , which bring together to 70 @-@ year @-@ born , are being put on the back in the heads of the two .
[2020-01-25 05:58:27] Total translation time: 147.54811s
[2020-01-25 05:58:28] [valid] Ep. 13 : Up. 150000 : translation : 15.43 : stalled 7 times (last best: 16.49)
[2020-01-25 05:58:28] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-25 05:58:32] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-25 06:08:25] Ep. 13 : Up. 151000 : Sen. 268,509 : Cost 18.42560577 : Time 759.58s : 1596.21 words/s
[2020-01-25 06:18:06] Ep. 13 : Up. 152000 : Sen. 317,080 : Cost 18.50701332 : Time 581.00s : 2084.14 words/s
[2020-01-25 06:27:44] Ep. 13 : Up. 153000 : Sen. 365,796 : Cost 18.53766251 : Time 578.12s : 2093.34 words/s
[2020-01-25 06:37:22] Ep. 13 : Up. 154000 : Sen. 414,530 : Cost 18.60911560 : Time 578.27s : 2093.31 words/s
[2020-01-25 06:47:02] Ep. 13 : Up. 155000 : Sen. 463,128 : Cost 18.90910530 : Time 579.81s : 2088.36 words/s
[2020-01-25 06:56:45] Ep. 13 : Up. 156000 : Sen. 512,039 : Cost 19.07496834 : Time 583.17s : 2097.20 words/s
[2020-01-25 07:06:22] Ep. 13 : Up. 157000 : Sen. 560,396 : Cost 19.12592888 : Time 576.85s : 2088.47 words/s
[2020-01-25 07:12:14] Seen 589960 samples
[2020-01-25 07:12:14] Starting epoch 14
[2020-01-25 07:12:14] [data] Shuffling data
[2020-01-25 07:12:14] [data] Done reading 640633 sentences
[2020-01-25 07:12:17] [data] Done shuffling 640633 sentences to temp files
[2020-01-25 07:16:04] Ep. 14 : Up. 158000 : Sen. 19,065 : Cost 17.84132576 : Time 581.93s : 2076.09 words/s
[2020-01-25 07:25:43] Ep. 14 : Up. 159000 : Sen. 67,448 : Cost 16.16341209 : Time 579.40s : 2093.45 words/s
[2020-01-25 07:35:19] Ep. 14 : Up. 160000 : Sen. 116,369 : Cost 16.11441040 : Time 575.87s : 2102.59 words/s
[2020-01-25 07:35:33] [valid] Ep. 14 : Up. 160000 : cross-entropy : 93.3989 : stalled 9 times (last best: 82.2896)
[2020-01-25 07:35:33] Translating validation set...
[2020-01-25 07:35:44] Best translation 0 : the strategy of the Republican Party against bilingual time
[2020-01-25 07:35:44] Best translation 1 : the leaders will see their policy necessary to fight electoral fraud .
[2020-01-25 07:35:44] Best translation 2 : however , the centre of a scam claims such a con as a myth that electoral fraud in the USA is less frequent than that of persons dying after the intervention .
[2020-01-25 07:35:44] Best translation 3 : after all , the last ten years have seen only 300 cases of electoral fraud in the US .
[2020-01-25 07:35:44] Best translation 4 : one thing is certain : these new provisions will have a negative impact on voter turnout .
[2020-01-25 07:35:44] Best translation 5 : in this sense , the measure is partly the American democratic system .
[2020-01-25 07:35:44] Best translation 10 : new electoral laws require the voters to present the identity of a photograph and Treaty of State citizenship .
[2020-01-25 07:35:44] Best translation 20 : in addition , the acquisition required claim is often more than a hundred dollars .
[2020-01-25 07:35:44] Best translation 40 : on the other hand , I think that there were only 46 % of the voters who voted in advance on behalf of 76 % of the voters .
[2020-01-25 07:35:44] Best translation 80 : cigarettes are responsible for 85 % of cases of lung cancer .
[2020-01-25 07:35:44] Best translation 160 : 'in life , people learn that they were not standing by a standing standing , ' Small Pierre & D , the passion in the use of Annexes .
[2020-01-25 07:35:54] Best translation 320 : this means that the screen is very powerful and that he makes a lot of frustration .
[2020-01-25 07:36:04] Best translation 640 : the President of Russia is proposing a tightening of the law in relation to the owners of civil weapons .
[2020-01-25 07:36:39] Best translation 1280 : in writing . - ( DE ) Unfortunately , just as in the Czech Republic the current society is based on the reckless benefit of their own good , regardless of the interest shown .
[2020-01-25 07:37:47] Best translation 2560 : once the large sulphur blocks of the sulphur , which combined with a 70 @-@ year @-@ old love , are on the back of the shoulders of the two @-@ year @-@ old ultra @-@ old @-@ old @-@ old shop .
[2020-01-25 07:38:01] Total translation time: 147.83862s
[2020-01-25 07:38:02] [valid] Ep. 14 : Up. 160000 : translation : 15.24 : stalled 8 times (last best: 16.49)
[2020-01-25 07:38:02] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-25 07:38:07] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-25 07:47:59] Ep. 14 : Up. 161000 : Sen. 165,195 : Cost 16.44795799 : Time 759.62s : 1598.23 words/s
[2020-01-25 07:57:39] Ep. 14 : Up. 162000 : Sen. 213,917 : Cost 16.65507317 : Time 580.05s : 2088.74 words/s
[2020-01-25 08:07:21] Ep. 14 : Up. 163000 : Sen. 262,805 : Cost 16.87364006 : Time 581.96s : 2093.77 words/s
[2020-01-25 08:17:00] Ep. 14 : Up. 164000 : Sen. 311,342 : Cost 16.99367332 : Time 579.79s : 2083.95 words/s
[2020-01-25 08:26:41] Ep. 14 : Up. 165000 : Sen. 360,238 : Cost 17.14024734 : Time 580.63s : 2092.88 words/s
[2020-01-25 08:36:18] Ep. 14 : Up. 166000 : Sen. 408,770 : Cost 17.30529022 : Time 577.31s : 2092.39 words/s
[2020-01-25 08:45:55] Ep. 14 : Up. 167000 : Sen. 457,473 : Cost 17.36047173 : Time 576.37s : 2094.26 words/s
[2020-01-25 08:55:31] Ep. 14 : Up. 168000 : Sen. 506,090 : Cost 17.57466888 : Time 576.73s : 2089.48 words/s
[2020-01-25 09:05:10] Ep. 14 : Up. 169000 : Sen. 554,502 : Cost 17.79125595 : Time 578.75s : 2090.95 words/s
[2020-01-25 09:12:14] Seen 589960 samples
[2020-01-25 09:12:14] Starting epoch 15
[2020-01-25 09:12:14] [data] Shuffling data
[2020-01-25 09:12:14] [data] Done reading 640633 sentences
[2020-01-25 09:12:17] [data] Done shuffling 640633 sentences to temp files
[2020-01-25 09:14:54] Ep. 15 : Up. 170000 : Sen. 13,154 : Cost 16.92607307 : Time 583.84s : 2072.91 words/s
[2020-01-25 09:15:09] [valid] Ep. 15 : Up. 170000 : cross-entropy : 94.6314 : stalled 10 times (last best: 82.2896)
[2020-01-25 09:15:09] Translating validation set...
[2020-01-25 09:15:19] Best translation 0 : the Republican Party strategy against bilingual re @-@ election
[2020-01-25 09:15:19] Best translation 1 : the leaders will see their policy necessary to combat electoral fraud .
[2020-01-25 09:15:19] Best translation 2 : however , the rebel Centre claims such a con for the myth that electoral fraud in the USA has been less frequent than the number of people who have died of oppression .
[2020-01-25 09:15:19] Best translation 3 : after all , the last ten years have seen only 300 cases of electoral fraud in the US .
[2020-01-25 09:15:19] Best translation 4 : one thing is certain : these new provisions will have a negative effect on voter turnout .
[2020-01-25 09:15:19] Best translation 5 : this measure partially distorts the American democratic system in this regard .
[2020-01-25 09:15:19] Best translation 10 : new electoral laws require voters to submit identity cards and evidence of US citizenship .
[2020-01-25 09:15:19] Best translation 20 : in addition , the purchase required claim is often more than a hundred dollars .
[2020-01-25 09:15:20] Best translation 40 : on the other side , entrepreneurs who present 76 % of the voters were only 46 % of the voters voted in advance .
[2020-01-25 09:15:20] Best translation 80 : cigarettes stand for 85 % of the cases of lung cancer .
[2020-01-25 09:15:20] Best translation 160 : assert throughout life , man is learning a standing ovation , ' Pierre Morel ' , in the case @-@ up , instance .
[2020-01-25 09:15:29] Best translation 320 : that means that equipment is very powerful and he has done a great deal .
[2020-01-25 09:15:40] Best translation 640 : for its part , Russia is proposing a tightening of the law in relation to the owners of civil weapons .
[2020-01-25 09:16:15] Best translation 1280 : in writing . - ( RO ) A final aspect is that the current society is based on the unscrupulous vested interests , regardless of the anti @-@ Europeans interest .
[2020-01-25 09:17:22] Best translation 2560 : after the heavy blocks of sulphur , which give a total of 70 @-@ year @-@ olds combined , bring them to the shoulders of the two .
[2020-01-25 09:17:36] Total translation time: 147.81091s
[2020-01-25 09:17:37] [valid] Ep. 15 : Up. 170000 : translation : 15.18 : stalled 9 times (last best: 16.49)
[2020-01-25 09:17:37] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-25 09:17:42] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz
[2020-01-25 09:17:53] Training finished
[2020-01-25 09:17:53] Saving model weights and runtime parameters to preprocess-train/model/model.npz
[2020-01-25 09:17:57] Saving Adam parameters to preprocess-train/model/model.npz.optimizer.npz