Hugging Face's logo

twieland
/
SCRATCH_ja-en_helsinki

Text2Text Generation
Transformers
PyTorch
marian
Generated from Trainer
Inference Endpoints
Model card Files Community
SCRATCH_ja-en_helsinki / README.md
twieland's picture
twieland
Update README.md
c90aedf
preview code
|
raw
history blame contribute delete
8.43 kB
metadata
license: apache-2.0
tags:
  - generated_from_trainer
model-index:
  - name: SCRATCH_ja-en_helsinki
    results: []

SCRATCH_ja-en_helsinki

This model is a fine-tuned version of Helsinki-NLP/opus-mt-ja-en on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 1.5583
  • Otaku Benchmark VN BLEU: 19.12
  • Otaku Benchmark LN BLEU: 11.55
  • Otaku Benchmark MANGA BLEU: 12.98

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.0003
  • train_batch_size: 96
  • eval_batch_size: 64
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 2
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss
3.0252 0.02 2000 2.4140
2.8406 0.03 4000 2.2819
2.7505 0.05 6000 2.3018
2.6948 0.06 8000 2.1931
2.6408 0.08 10000 2.1724
2.6004 0.09 12000 2.1583
2.5685 0.11 14000 2.1203
2.5432 0.12 16000 2.1593
2.5153 0.14 18000 2.1009
2.4906 0.15 20000 2.0899
2.4709 0.17 22000 2.0512
2.4471 0.18 24000 2.0208
2.4295 0.2 26000 2.0773
2.4154 0.21 28000 2.0441
2.4008 0.23 30000 2.0235
2.3834 0.24 32000 2.0190
2.3709 0.26 34000 1.9831
2.3537 0.27 36000 1.9870
2.3486 0.29 38000 1.9692
2.3346 0.3 40000 1.9517
2.3195 0.32 42000 1.9800
2.3104 0.33 44000 1.9676
2.298 0.35 46000 1.9563
2.2905 0.36 48000 1.9217
2.2792 0.38 50000 1.9195
2.2714 0.39 52000 1.9109
2.2593 0.41 54000 1.9044
2.2582 0.42 56000 1.8876
2.2482 0.44 58000 1.8860
2.2394 0.45 60000 1.8887
2.2273 0.47 62000 1.8862
2.2255 0.48 64000 1.8705
2.2166 0.5 66000 1.8696
2.2075 0.51 68000 1.8657
2.1992 0.53 70000 1.8585
2.1969 0.54 72000 1.8526
2.1894 0.56 74000 1.8493
2.1817 0.57 76000 1.8480
2.1771 0.59 78000 1.8333
2.1683 0.6 80000 1.8342
2.1667 0.62 82000 1.8537
2.1546 0.63 84000 1.8261
2.1467 0.65 86000 1.8092
2.1421 0.66 88000 1.8137
2.1395 0.68 90000 1.8286
2.1313 0.69 92000 1.8042
2.1241 0.71 94000 1.7934
2.1214 0.72 96000 1.7940
2.12 0.74 98000 1.8064
2.1096 0.75 100000 1.7983
2.1035 0.77 102000 1.8089
2.0937 0.78 104000 1.7941
2.0893 0.8 106000 1.7791
2.0869 0.81 108000 1.7807
2.0845 0.83 110000 1.7852
2.0782 0.84 112000 1.7675
2.0755 0.86 114000 1.7756
2.0657 0.87 116000 1.7604
2.0614 0.89 118000 1.7447
2.0591 0.9 120000 1.7489
2.0586 0.92 122000 1.7550
2.0498 0.93 124000 1.7543
2.0455 0.95 126000 1.7510
2.04 0.96 128000 1.7439
2.0385 0.98 130000 1.7407
2.0267 0.99 132000 1.7467
2.0088 1.01 134000 1.7455
1.9826 1.02 136000 1.7210
1.9785 1.04 138000 1.7524
1.9777 1.05 140000 1.7272
1.9763 1.07 142000 1.7283
1.9736 1.08 144000 1.7210
1.9704 1.1 146000 1.7001
1.9625 1.11 148000 1.7112
1.9665 1.13 150000 1.7236
1.9592 1.14 152000 1.7169
1.9606 1.16 154000 1.6962
1.9571 1.17 156000 1.7064
1.9532 1.19 158000 1.6898
1.9465 1.2 160000 1.7004
1.9438 1.22 162000 1.7092
1.9435 1.23 164000 1.6927
1.9361 1.25 166000 1.6838
1.9369 1.26 168000 1.6784
1.9287 1.28 170000 1.6709
1.928 1.29 172000 1.6735
1.9227 1.31 174000 1.6689
1.9213 1.32 176000 1.6685
1.9152 1.34 178000 1.6635
1.9092 1.35 180000 1.6561
1.9059 1.37 182000 1.6673
1.9094 1.38 184000 1.6717
1.9006 1.4 186000 1.6593
1.8956 1.41 188000 1.6483
1.8972 1.43 190000 1.6635
1.8907 1.44 192000 1.6604
1.8885 1.46 194000 1.6465
1.8844 1.47 196000 1.6444
1.8799 1.49 198000 1.6307
1.8813 1.5 200000 1.6240
1.8693 1.52 202000 1.6102
1.8768 1.53 204000 1.6197
1.8678 1.55 206000 1.6275
1.8588 1.56 208000 1.6183
1.8585 1.58 210000 1.6197
1.8564 1.59 212000 1.6004
1.8493 1.61 214000 1.6078
1.85 1.62 216000 1.6001
1.8428 1.64 218000 1.6106
1.8428 1.65 220000 1.5866
1.8423 1.67 222000 1.5993
1.8352 1.68 224000 1.6052
1.8385 1.7 226000 1.5959
1.8307 1.71 228000 1.6024
1.8248 1.73 230000 1.5969
1.82 1.74 232000 1.5878
1.8254 1.76 234000 1.5934
1.8188 1.77 236000 1.5827
1.813 1.79 238000 1.5797
1.8128 1.8 240000 1.5758
1.8044 1.82 242000 1.5752
1.808 1.83 244000 1.5818
1.8025 1.85 246000 1.5772
1.7992 1.86 248000 1.5738
1.8021 1.88 250000 1.5752
1.7988 1.89 252000 1.5717
1.7967 1.91 254000 1.5690
1.7909 1.92 256000 1.5607
1.7942 1.94 258000 1.5618
1.7897 1.95 260000 1.5585
1.7871 1.97 262000 1.5576
1.7843 1.98 264000 1.5577
1.7888 2.0 266000 1.5583

Framework versions

  • Transformers 4.19.2
  • Pytorch 1.11.0+cu113
  • Datasets 2.2.2
  • Tokenizers 0.12.1