SCRATCH_ja-en_helsinki

twieland/SCRATCH_ja-en_helsinki

SCRATCH_ja-en_helsinki is a machine learning model from twieland. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for transformers. The card lists the license as apache-2.0.

should probably proofread and complete it, then remove this comment. --

Downloads · 30 days

2% of all-time downloads

All-time downloads

503

Public

Repo size

40.1 GB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.bin301 MB · 99%

At a glance

Library: transformers
License: apache-2.0
Model type: marian
Access: Public
Created: Jun 11, 2022
Updated: Jun 11, 2022
SHA: c90aedfd

Model card

From the Hugging Face model README

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

Library: transformers
Type: marian
License: apache-2.0
Created: Jun 11, 2022
Updated: Jun 11, 2022