7ea07d0f69a1e5485200faa2f305f4f2

This model is a fine-tuned version of google/umt5-base on the Helsinki-NLP/opus_books [fr-sv] dataset. It achieves the following results on the evaluation set:

• Loss: 2.8795
• Data Size: 1.0
• Epoch Runtime: 21.3988
• Bleu: 4.4021

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	12.6891	0	2.1175	0.0119
No log	1	75	12.6997	0.0078	2.4211	0.0247
No log	2	150	12.4566	0.0156	3.0919	0.0122
No log	3	225	12.4297	0.0312	4.4245	0.0402
No log	4	300	11.7860	0.0625	5.3923	0.0366
No log	5	375	11.5196	0.125	7.3997	0.0333
No log	6	450	10.7898	0.25	9.9374	0.0289
No log	7	525	8.5665	0.5	13.8185	0.0768
8.4747	8.0	600	5.9402	1.0	22.1808	0.4226
6.5483	9.0	675	4.3072	1.0	20.1638	2.2010
5.0775	10.0	750	3.6728	1.0	20.1892	4.4350
4.6807	11.0	825	3.3535	1.0	21.4876	5.0813
4.2284	12.0	900	3.2363	1.0	20.5669	2.7329
4.0264	13.0	975	3.1474	1.0	20.7894	2.9965
3.8285	14.0	1050	3.0936	1.0	20.9474	3.1947
3.6863	15.0	1125	3.0358	1.0	20.6891	3.4564
3.5269	16.0	1200	3.0083	1.0	21.6643	3.4921
3.4742	17.0	1275	2.9808	1.0	20.0551	3.6284
3.382	18.0	1350	2.9654	1.0	20.7244	3.6758
3.2942	19.0	1425	2.9437	1.0	21.6724	3.8500
3.2333	20.0	1500	2.9292	1.0	21.4422	3.8753
3.1626	21.0	1575	2.9188	1.0	20.4945	3.8748
3.1052	22.0	1650	2.9139	1.0	21.0036	3.9785
3.0291	23.0	1725	2.9109	1.0	21.3788	4.0008
2.9656	24.0	1800	2.9018	1.0	21.3974	4.0299
2.9288	25.0	1875	2.8930	1.0	20.7319	4.1182
2.8649	26.0	1950	2.8929	1.0	20.2559	4.1961
2.8202	27.0	2025	2.8883	1.0	20.3763	4.2134
2.7626	28.0	2100	2.8825	1.0	20.6661	4.3136
2.6921	29.0	2175	2.8747	1.0	21.0836	4.2935
2.6788	30.0	2250	2.8861	1.0	22.7795	4.2313
2.6414	31.0	2325	2.8776	1.0	19.8240	4.3317
2.605	32.0	2400	2.8784	1.0	20.4296	4.3686
2.5408	33.0	2475	2.8795	1.0	21.3988	4.4021

Framework versions

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