c72031b135eef954fc0610e6d8bc219e

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

• Loss: 2.1574
• Data Size: 1.0
• Epoch Runtime: 71.1753
• Bleu: 7.5301

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	16.9728	0	6.3246	0.2711
No log	1	437	16.2460	0.0078	6.9613	0.3604
No log	2	874	15.0040	0.0156	7.7570	0.4340
No log	3	1311	13.1679	0.0312	9.0722	0.4471
No log	4	1748	10.7075	0.0625	10.9963	0.5078
11.9865	5	2185	7.3649	0.125	15.0259	0.6189
8.482	6	2622	4.9792	0.25	22.9080	1.0091
5.5888	7	3059	3.8803	0.5	37.6092	3.9858
4.4134	8.0	3496	3.1619	1.0	69.0978	3.0325
3.9851	9.0	3933	2.8942	1.0	69.1926	3.8939
3.701	10.0	4370	2.7730	1.0	68.8371	4.2735
3.5848	11.0	4807	2.6992	1.0	68.6402	4.6192
3.4352	12.0	5244	2.6427	1.0	69.9948	4.8753
3.3044	13.0	5681	2.5969	1.0	69.2732	5.0178
3.2305	14.0	6118	2.5579	1.0	70.2969	5.1887
3.1587	15.0	6555	2.5228	1.0	69.0757	5.3367
3.0715	16.0	6992	2.4845	1.0	70.2242	5.5283
3.0601	17.0	7429	2.4683	1.0	69.8976	5.6228
2.9926	18.0	7866	2.4341	1.0	69.7541	5.8032
2.9095	19.0	8303	2.4180	1.0	69.9274	5.9151
2.9047	20.0	8740	2.4011	1.0	70.2386	5.9841
2.8405	21.0	9177	2.3805	1.0	69.6005	6.0747
2.7902	22.0	9614	2.3596	1.0	70.0786	6.1834
2.7556	23.0	10051	2.3500	1.0	70.0480	6.2457
2.7298	24.0	10488	2.3324	1.0	69.6931	6.3249
2.7325	25.0	10925	2.3186	1.0	72.4412	6.4368
2.689	26.0	11362	2.3087	1.0	69.4702	6.4923
2.6729	27.0	11799	2.3005	1.0	70.1461	6.5329
2.5953	28.0	12236	2.2843	1.0	70.2789	6.6074
2.5617	29.0	12673	2.2735	1.0	70.2161	6.6736
2.531	30.0	13110	2.2674	1.0	69.6573	6.6912
2.5226	31.0	13547	2.2657	1.0	70.6520	6.7476
2.5702	32.0	13984	2.2580	1.0	70.1485	6.8362
2.5027	33.0	14421	2.2420	1.0	70.7487	6.8910
2.483	34.0	14858	2.2354	1.0	70.1335	6.8995
2.4384	35.0	15295	2.2254	1.0	70.7653	7.0147
2.416	36.0	15732	2.2219	1.0	70.0677	7.0343
2.3831	37.0	16169	2.2098	1.0	70.6963	7.0994
2.3674	38.0	16606	2.2080	1.0	70.6031	7.0817
2.3575	39.0	17043	2.2006	1.0	70.4574	7.1965
2.3313	40.0	17480	2.1937	1.0	70.7618	7.1884
2.3405	41.0	17917	2.1935	1.0	69.9549	7.2435
2.2925	42.0	18354	2.1881	1.0	70.2700	7.2781
2.298	43.0	18791	2.1820	1.0	69.4775	7.3183
2.2471	44.0	19228	2.1749	1.0	70.4325	7.3587
2.2876	45.0	19665	2.1698	1.0	70.3586	7.4095
2.2149	46.0	20102	2.1743	1.0	70.6157	7.4523
2.2064	47.0	20539	2.1638	1.0	69.7505	7.4806
2.1987	48.0	20976	2.1674	1.0	70.9113	7.4668
2.1447	49.0	21413	2.1589	1.0	70.1095	7.5387
2.1704	50.0	21850	2.1574	1.0	71.1753	7.5301

Framework versions

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