MIX2_en-ja_helsinki

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

• Loss: 1.6703

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: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
3.5357	0.02	4000	2.9519
2.8601	0.04	8000	2.6962
2.6183	0.06	12000	2.5156
2.4731	0.08	16000	2.4312
2.3731	0.1	20000	2.3575
2.2964	0.11	24000	2.3319
2.238	0.13	28000	2.2802
2.1919	0.15	32000	2.2552
2.1479	0.17	36000	2.2354
2.1104	0.19	40000	2.2210
2.0788	0.21	44000	2.1835
2.0552	0.23	48000	2.1391
2.0228	0.25	52000	2.1338
2.0062	0.27	56000	2.1115
1.9868	0.29	60000	2.1025
1.9628	0.31	64000	2.1334
1.9474	0.32	68000	2.0935
1.9318	0.34	72000	2.1030
1.9187	0.36	76000	2.0605
1.9019	0.38	80000	2.0388
1.8916	0.4	84000	2.0360
1.8775	0.42	88000	2.0356
1.8689	0.44	92000	2.0315
1.8558	0.46	96000	2.0169
1.8431	0.48	100000	2.0213
1.8373	0.5	104000	2.0071
1.8224	0.52	108000	2.0093
1.8181	0.53	112000	1.9952
1.8087	0.55	116000	1.9927
1.7998	0.57	120000	1.9726
1.7947	0.59	124000	1.9817
1.7874	0.61	128000	1.9650
1.7781	0.63	132000	1.9688
1.7712	0.65	136000	1.9655
1.7631	0.67	140000	1.9561
1.7577	0.69	144000	1.9529
1.7528	0.71	148000	1.9447
1.746	0.73	152000	1.9700
1.7386	0.74	156000	1.9413
1.7329	0.76	160000	1.9329
1.7285	0.78	164000	1.9289
1.7227	0.8	168000	1.9337
1.7186	0.82	172000	1.9263
1.7116	0.84	176000	1.9407
1.7072	0.86	180000	1.9059
1.7032	0.88	184000	1.9380
1.6932	0.9	188000	1.9183
1.6921	0.92	192000	1.9131
1.6875	0.94	196000	1.9180
1.6846	0.96	200000	1.9040
1.6797	0.97	204000	1.9089
1.6725	0.99	208000	1.9024
1.6589	1.01	212000	1.8909
1.6507	1.03	216000	1.8837
1.6441	1.05	220000	1.8906
1.6445	1.07	224000	1.8914
1.6394	1.09	228000	1.8833
1.6382	1.11	232000	1.8837
1.6376	1.13	236000	1.8869
1.6329	1.15	240000	1.8829
1.6294	1.17	244000	1.8845
1.6273	1.18	248000	1.8888
1.6243	1.2	252000	1.8709
1.6226	1.22	256000	1.8418
1.6177	1.24	260000	1.8587
1.6151	1.26	264000	1.8526
1.6111	1.28	268000	1.8494
1.6084	1.3	272000	1.8781
1.6043	1.32	276000	1.8390
1.6011	1.34	280000	1.8603
1.5999	1.36	284000	1.8515
1.5954	1.38	288000	1.8356
1.5936	1.39	292000	1.8530
1.5916	1.41	296000	1.8475
1.5886	1.43	300000	1.8410
1.5883	1.45	304000	1.8153
1.5828	1.47	308000	1.8254
1.582	1.49	312000	1.8139
1.578	1.51	316000	1.8366
1.5723	1.53	320000	1.8353
1.5705	1.55	324000	1.8230
1.5691	1.57	328000	1.8194
1.5656	1.59	332000	1.8069
1.566	1.6	336000	1.8204
1.5604	1.62	340000	1.8307
1.5573	1.64	344000	1.8209
1.5547	1.66	348000	1.8320
1.5545	1.68	352000	1.8179
1.5519	1.7	356000	1.8323
1.545	1.72	360000	1.8005
1.5483	1.74	364000	1.8034
1.5454	1.76	368000	1.7997
1.5393	1.78	372000	1.8078
1.5381	1.8	376000	1.8204
1.5347	1.81	380000	1.8071
1.5327	1.83	384000	1.7997
1.529	1.85	388000	1.8012
1.5287	1.87	392000	1.8028
1.5273	1.89	396000	1.8103
1.5194	1.91	400000	1.8008
1.5197	1.93	404000	1.8004
1.5218	1.95	408000	1.8024
1.514	1.97	412000	1.7852
1.5146	1.99	416000	1.7908
1.5045	2.01	420000	1.7864
1.4876	2.02	424000	1.7813
1.4846	2.04	428000	1.7822
1.4865	2.06	432000	1.7737
1.4857	2.08	436000	1.7668
1.4825	2.1	440000	1.7681
1.4828	2.12	444000	1.7685
1.4821	2.14	448000	1.7636
1.4778	2.16	452000	1.7778
1.4803	2.18	456000	1.7834
1.4766	2.2	460000	1.7801
1.4741	2.22	464000	1.7601
1.4705	2.23	468000	1.7665
1.4739	2.25	472000	1.7604
1.4694	2.27	476000	1.7803
1.4665	2.29	480000	1.7835
1.4668	2.31	484000	1.7670
1.4605	2.33	488000	1.7629
1.4626	2.35	492000	1.7612
1.4627	2.37	496000	1.7612
1.4569	2.39	500000	1.7557
1.455	2.41	504000	1.7599
1.4547	2.43	508000	1.7569
1.453	2.44	512000	1.7589
1.4515	2.46	516000	1.7679
1.4501	2.48	520000	1.7574
1.4446	2.5	524000	1.7526
1.4456	2.52	528000	1.7506
1.4445	2.54	532000	1.7484
1.4428	2.56	536000	1.7447
1.439	2.58	540000	1.7468
1.441	2.6	544000	1.7609
1.4358	2.62	548000	1.7498
1.4318	2.64	552000	1.7592
1.4276	2.65	556000	1.7452
1.4317	2.67	560000	1.7500
1.4277	2.69	564000	1.7392
1.4259	2.71	568000	1.7351
1.4239	2.73	572000	1.7385
1.4191	2.75	576000	1.7487
1.4204	2.77	580000	1.7392
1.4176	2.79	584000	1.7372
1.4147	2.81	588000	1.7347
1.4154	2.83	592000	1.7085
1.4134	2.85	596000	1.7103
1.4091	2.87	600000	1.7124
1.4091	2.88	604000	1.7369
1.406	2.9	608000	1.7142
1.4028	2.92	612000	1.7376
1.4019	2.94	616000	1.7201
1.4018	2.96	620000	1.7230
1.3959	2.98	624000	1.7206
1.3985	3.0	628000	1.7183
1.3681	3.02	632000	1.7283
1.3668	3.04	636000	1.7330
1.3687	3.06	640000	1.7187
1.3681	3.08	644000	1.7163
1.3687	3.09	648000	1.7249
1.364	3.11	652000	1.7283
1.364	3.13	656000	1.7091
1.3652	3.15	660000	1.7030
1.3623	3.17	664000	1.7058
1.3604	3.19	668000	1.7101
1.3598	3.21	672000	1.7104
1.3577	3.23	676000	1.7028
1.3574	3.25	680000	1.7023
1.3546	3.27	684000	1.7197
1.3549	3.29	688000	1.7045
1.3534	3.3	692000	1.6990
1.3511	3.32	696000	1.6971
1.3504	3.34	700000	1.6894
1.346	3.36	704000	1.6820
1.3467	3.38	708000	1.6920
1.3461	3.4	712000	1.6897
1.3425	3.42	716000	1.6962
1.34	3.44	720000	1.6864
1.3408	3.46	724000	1.6860
1.3387	3.48	728000	1.6924
1.3377	3.5	732000	1.6919
1.3378	3.51	736000	1.6858
1.334	3.53	740000	1.6816
1.3347	3.55	744000	1.6867
1.3307	3.57	748000	1.6859
1.3316	3.59	752000	1.6896
1.3257	3.61	756000	1.6824
1.3222	3.63	760000	1.6819
1.3247	3.65	764000	1.6809
1.3207	3.67	768000	1.6775
1.3227	3.69	772000	1.6807
1.3203	3.71	776000	1.6750
1.3203	3.72	780000	1.6758
1.316	3.74	784000	1.6787
1.3147	3.76	788000	1.6747
1.3146	3.78	792000	1.6718
1.3137	3.8	796000	1.6744
1.3143	3.82	800000	1.6733
1.3123	3.84	804000	1.6754
1.3069	3.86	808000	1.6734
1.3122	3.88	812000	1.6742
1.3074	3.9	816000	1.6742
1.3006	3.92	820000	1.6709
1.308	3.93	824000	1.6714
1.3063	3.95	828000	1.6727
1.3036	3.97	832000	1.6711
1.3048	3.99	836000	1.6703

Framework versions

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