En-Nso_update

This model is a fine-tuned version of Helsinki-NLP/opus-mt-en-nso on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.8782
• Bleu: 31.2967

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu
No log	1.0	4	7.2950	0.0088
No log	2.0	8	5.9614	0.6848
No log	3.0	12	5.0695	4.9050
No log	4.0	16	4.5523	9.1757
No log	5.0	20	4.2355	10.4744
No log	6.0	24	4.0106	14.6163
No log	7.0	28	3.8427	15.8379
No log	8.0	32	3.7264	15.6158
No log	9.0	36	3.6338	16.3562
No log	10.0	40	3.5555	21.1011
No log	11.0	44	3.4839	21.5754
No log	12.0	48	3.4180	22.7155
No log	13.0	52	3.3620	23.1592
No log	14.0	56	3.3115	24.3886
No log	15.0	60	3.2676	24.1278
No log	16.0	64	3.2285	24.2245
No log	17.0	68	3.1974	23.9716
No log	18.0	72	3.1695	24.2395
No log	19.0	76	3.1441	23.3442
No log	20.0	80	3.1235	21.3332
No log	21.0	84	3.1029	21.8410
No log	22.0	88	3.0849	22.4065
No log	23.0	92	3.0666	22.3016
No log	24.0	96	3.0534	22.9616
No log	25.0	100	3.0423	23.3971
No log	26.0	104	3.0306	23.5443
No log	27.0	108	3.0183	23.3348
No log	28.0	112	3.0051	23.4077
No log	29.0	116	2.9947	24.1791
No log	30.0	120	2.9855	24.1265
No log	31.0	124	2.9777	23.9860
No log	32.0	128	2.9691	24.7301
No log	33.0	132	2.9597	25.1896
No log	34.0	136	2.9521	24.5893
No log	35.0	140	2.9457	24.5229
No log	36.0	144	2.9409	24.6232
No log	37.0	148	2.9354	24.2830
No log	38.0	152	2.9322	26.1404
No log	39.0	156	2.9306	25.9425
No log	40.0	160	2.9288	30.5432
No log	41.0	164	2.9261	29.4635
No log	42.0	168	2.9215	28.4787
No log	43.0	172	2.9182	28.9082
No log	44.0	176	2.9151	29.3171
No log	45.0	180	2.9132	28.3602
No log	46.0	184	2.9126	28.9583
No log	47.0	188	2.9104	26.0269
No log	48.0	192	2.9086	29.6904
No log	49.0	196	2.9052	29.2881
No log	50.0	200	2.9020	29.6063
No log	51.0	204	2.8994	29.5224
No log	52.0	208	2.8960	29.3913
No log	53.0	212	2.8930	30.5451
No log	54.0	216	2.8889	32.1862
No log	55.0	220	2.8869	31.9423
No log	56.0	224	2.8859	30.7244
No log	57.0	228	2.8846	30.8172
No log	58.0	232	2.8837	30.5376
No log	59.0	236	2.8826	31.1454
No log	60.0	240	2.8813	30.9049
No log	61.0	244	2.8802	30.6363
No log	62.0	248	2.8802	31.3739
No log	63.0	252	2.8799	30.9776
No log	64.0	256	2.8793	29.8283
No log	65.0	260	2.8795	29.6912
No log	66.0	264	2.8804	29.7654
No log	67.0	268	2.8810	29.1586
No log	68.0	272	2.8822	28.8888
No log	69.0	276	2.8819	29.7222
No log	70.0	280	2.8810	29.9932
No log	71.0	284	2.8811	30.2492
No log	72.0	288	2.8802	29.9644
No log	73.0	292	2.8791	30.3378
No log	74.0	296	2.8790	29.8055
No log	75.0	300	2.8794	29.0100
No log	76.0	304	2.8795	30.7968
No log	77.0	308	2.8790	31.5414
No log	78.0	312	2.8783	31.5060
No log	79.0	316	2.8775	31.4376
No log	80.0	320	2.8766	31.6005
No log	81.0	324	2.8767	31.3697
No log	82.0	328	2.8769	31.6108
No log	83.0	332	2.8770	31.4214
No log	84.0	336	2.8772	31.6039
No log	85.0	340	2.8776	32.0254
No log	86.0	344	2.8779	31.4024
No log	87.0	348	2.8783	32.0279
No log	88.0	352	2.8786	31.8914
No log	89.0	356	2.8788	31.6500
No log	90.0	360	2.8791	31.7698
No log	91.0	364	2.8793	31.6137
No log	92.0	368	2.8793	31.8244
No log	93.0	372	2.8790	31.5626
No log	94.0	376	2.8786	31.3743
No log	95.0	380	2.8785	31.4160
No log	96.0	384	2.8784	31.6682
No log	97.0	388	2.8782	31.8335
No log	98.0	392	2.8782	31.7143
No log	99.0	396	2.8782	31.7143
No log	100.0	400	2.8782	31.7143

Framework versions

• Transformers 4.20.1
• Pytorch 1.12.0+cu113
• Datasets 2.3.2
• Tokenizers 0.12.1