b3c9ef86b1d5d0d97a0a79dad148ceda

This model is a fine-tuned version of google-t5/t5-small on the Helsinki-NLP/opus_books dataset. It achieves the following results on the evaluation set:

• Loss: 2.5441
• Data Size: 1.0
• Epoch Runtime: 5.3293
• Bleu: 2.6579

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: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Data Size	Epoch Runtime	Bleu
No log	0	0	4.7305	0	1.2510	0.2816
No log	1	29	4.7180	0.0078	1.4901	0.2862
No log	2	58	4.6720	0.0156	1.4777	0.3111
No log	3	87	4.5907	0.0312	1.6292	0.3221
No log	4	116	4.5183	0.0625	1.6536	0.3421
No log	5	145	4.3042	0.125	1.9016	0.4274
0.4293	6	174	4.0675	0.25	2.3506	0.4365
0.4293	7	203	3.8627	0.5	3.0871	0.4041
0.4293	8.0	232	3.6116	1.0	4.8709	0.5755
2.6111	9.0	261	3.4559	1.0	4.8250	0.6727
2.6111	10.0	290	3.3505	1.0	5.1899	0.5885
3.6423	11.0	319	3.2728	1.0	5.0376	0.6166
3.6423	12.0	348	3.2053	1.0	5.6061	0.7122
3.4945	13.0	377	3.1481	1.0	5.5992	0.9659
3.3686	14.0	406	3.1029	1.0	5.1744	1.2183
3.3686	15.0	435	3.0584	1.0	5.4537	1.3744
3.2785	16.0	464	3.0217	1.0	5.5987	1.3840
3.2785	17.0	493	2.9864	1.0	5.8839	1.5402
3.2022	18.0	522	2.9581	1.0	5.5452	1.6641
3.1321	19.0	551	2.9299	1.0	5.3498	1.7430
3.1321	20.0	580	2.9091	1.0	5.4108	1.7930
3.0773	21.0	609	2.8837	1.0	6.0165	1.8715
3.0773	22.0	638	2.8596	1.0	5.0093	1.9125
3.0334	23.0	667	2.8459	1.0	5.4066	1.9491
3.0334	24.0	696	2.8206	1.0	5.2901	1.9523
2.9651	25.0	725	2.8017	1.0	5.4400	2.1171
2.9363	26.0	754	2.7851	1.0	5.2300	2.1637
2.9363	27.0	783	2.7704	1.0	4.9590	2.1843
2.8841	28.0	812	2.7528	1.0	4.9944	2.2392
2.8841	29.0	841	2.7386	1.0	5.0581	2.2154
2.8376	30.0	870	2.7235	1.0	4.8529	2.2586
2.8376	31.0	899	2.7104	1.0	5.0313	2.2268
2.7906	32.0	928	2.7006	1.0	5.3726	2.2246
2.7657	33.0	957	2.6877	1.0	5.5822	2.3319
2.7657	34.0	986	2.6763	1.0	6.0219	2.3137
2.7318	35.0	1015	2.6649	1.0	5.1100	2.3610
2.7318	36.0	1044	2.6565	1.0	5.4359	2.4039
2.703	37.0	1073	2.6445	1.0	5.3567	2.3845
2.661	38.0	1102	2.6377	1.0	5.3200	2.4204
2.661	39.0	1131	2.6294	1.0	4.8490	2.4322
2.6324	40.0	1160	2.6201	1.0	4.9322	2.5128
2.6324	41.0	1189	2.6127	1.0	5.2830	2.4622
2.5958	42.0	1218	2.6040	1.0	5.1769	2.5315
2.5958	43.0	1247	2.5959	1.0	5.2128	2.5383
2.5785	44.0	1276	2.5819	1.0	5.4421	2.5478
2.5474	45.0	1305	2.5794	1.0	5.2519	2.5611
2.5474	46.0	1334	2.5719	1.0	5.2863	2.5447
2.5187	47.0	1363	2.5648	1.0	5.3122	2.5498
2.5187	48.0	1392	2.5618	1.0	5.3430	2.5663
2.5019	49.0	1421	2.5574	1.0	5.3949	2.6240
2.4799	50.0	1450	2.5441	1.0	5.3293	2.6579

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.2.0
• Tokenizers 0.22.1