8b397e93122f02e251305ab6a7ea0137

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

• Loss: 1.7226
• Data Size: 1.0
• Epoch Runtime: 101.4265
• Bleu: 9.3907

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	11.3294	0	8.7264	0.3564
No log	1	434	11.2120	0.0078	9.6077	0.3557
No log	2	868	10.8344	0.0156	11.1388	0.4100
No log	3	1302	9.9545	0.0312	12.9233	0.3901
No log	4	1736	8.7990	0.0625	16.4405	0.5119
0.578	5	2170	9.4236	0.125	22.8667	0.2615
8.3328	6	2604	4.5183	0.25	33.2893	1.8934
4.0097	7	3038	2.5764	0.5	56.6631	4.8373
3.0065	8.0	3472	2.1856	1.0	101.8494	5.9452
2.676	9.0	3906	2.0421	1.0	101.0378	6.6706
2.5174	10.0	4340	1.9713	1.0	101.9740	7.0471
2.3989	11.0	4774	1.9171	1.0	102.9819	7.3988
2.2626	12.0	5208	1.8839	1.0	103.4988	7.6658
2.1977	13.0	5642	1.8490	1.0	103.9226	7.8602
2.1132	14.0	6076	1.8302	1.0	101.6970	8.0115
2.0603	15.0	6510	1.8026	1.0	101.7626	8.1996
2.006	16.0	6944	1.7834	1.0	102.9606	8.3058
1.9383	17.0	7378	1.7735	1.0	102.9428	8.4403
1.8985	18.0	7812	1.7613	1.0	103.7445	8.5071
1.8056	19.0	8246	1.7496	1.0	103.5564	8.6523
1.7996	20.0	8680	1.7423	1.0	102.8571	8.7177
1.7504	21.0	9114	1.7345	1.0	103.3688	8.8099
1.6857	22.0	9548	1.7290	1.0	102.4989	8.8871
1.6491	23.0	9982	1.7299	1.0	101.9331	8.9197
1.613	24.0	10416	1.7237	1.0	102.1754	8.9455
1.6032	25.0	10850	1.7213	1.0	103.9956	9.0970
1.5407	26.0	11284	1.7249	1.0	103.5786	9.1021
1.5496	27.0	11718	1.7145	1.0	102.7294	9.1801
1.4706	28.0	12152	1.7169	1.0	102.4893	9.2289
1.4444	29.0	12586	1.7156	1.0	102.2135	9.2741
1.4228	30.0	13020	1.7265	1.0	101.6165	9.3033
1.4021	31.0	13454	1.7226	1.0	101.4265	9.3907

Framework versions

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