119d40fde8cc3c24912b601104f9dd6b

This model is a fine-tuned version of google/umt5-small on the Helsinki-NLP/opus_books [fr-no] dataset. It achieves the following results on the evaluation set:

• Loss: 3.2158
• Data Size: 1.0
• Epoch Runtime: 15.7291
• Bleu: 3.7429

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	20.2441	0	1.8601	0.0704
No log	1	86	20.4134	0.0078	1.9973	0.0483
No log	2	172	20.2769	0.0156	3.0826	0.0462
No log	3	258	20.0506	0.0312	3.5037	0.0504
No log	4	344	19.3243	0.0625	4.1233	0.0462
1.0975	5	430	18.2588	0.125	4.8835	0.0467
4.6037	6	516	15.7345	0.25	6.3307	0.0446
5.5027	7	602	11.2304	0.5	9.4857	0.0542
7.0204	8.0	688	7.5414	1.0	15.6707	0.1923
8.4824	9.0	774	6.1525	1.0	14.3672	0.1869
6.9753	10.0	860	5.0391	1.0	14.2547	0.8677
6.4922	11.0	946	4.7010	1.0	14.1415	0.7576
5.8641	12.0	1032	4.4488	1.0	14.3490	0.8668
5.4654	13.0	1118	4.2347	1.0	14.3414	1.1044
5.1539	14.0	1204	4.0326	1.0	14.9090	1.4052
5.0033	15.0	1290	3.8507	1.0	14.8050	1.5826
4.7892	16.0	1376	3.7268	1.0	14.3627	1.7526
4.6346	17.0	1462	3.6404	1.0	14.7356	1.9106
4.5431	18.0	1548	3.5882	1.0	15.3488	2.1228
4.4212	19.0	1634	3.5567	1.0	15.1713	2.3188
4.3455	20.0	1720	3.5244	1.0	15.3461	2.4114
4.2834	21.0	1806	3.4975	1.0	15.7892	2.4112
4.2198	22.0	1892	3.4743	1.0	16.6737	2.4421
4.123	23.0	1978	3.4423	1.0	14.5820	2.6232
4.0733	24.0	2064	3.4133	1.0	14.3646	2.7541
3.9988	25.0	2150	3.3990	1.0	14.7306	2.8319
3.9794	26.0	2236	3.3789	1.0	15.2129	2.8298
3.9185	27.0	2322	3.3705	1.0	15.3848	2.8578
3.8666	28.0	2408	3.3515	1.0	15.2893	2.9725
3.8388	29.0	2494	3.3385	1.0	15.2697	3.0124
3.8084	30.0	2580	3.3300	1.0	15.5868	3.0930
3.7702	31.0	2666	3.3205	1.0	14.2415	3.0933
3.7115	32.0	2752	3.3123	1.0	14.3792	3.1280
3.6728	33.0	2838	3.3053	1.0	14.4939	3.1815
3.6494	34.0	2924	3.3076	1.0	14.8109	3.1974
3.6546	35.0	3010	3.2860	1.0	15.6599	3.2049
3.6139	36.0	3096	3.2780	1.0	15.4160	3.3345
3.5774	37.0	3182	3.2705	1.0	15.2976	3.3522
3.5437	38.0	3268	3.2703	1.0	15.9273	3.3657
3.4861	39.0	3354	3.2604	1.0	14.6697	3.3889
3.478	40.0	3440	3.2512	1.0	14.7191	3.3770
3.4634	41.0	3526	3.2473	1.0	15.4507	3.4208
3.4544	42.0	3612	3.2548	1.0	15.4143	3.5544
3.418	43.0	3698	3.2395	1.0	15.8460	3.5035
3.4076	44.0	3784	3.2340	1.0	15.6204	3.5561
3.3806	45.0	3870	3.2334	1.0	15.7972	3.6003
3.3351	46.0	3956	3.2240	1.0	14.5148	3.6058
3.3344	47.0	4042	3.2176	1.0	15.0169	3.5453
3.2764	48.0	4128	3.2136	1.0	15.1349	3.5881
3.2828	49.0	4214	3.2134	1.0	15.5410	3.6484
3.2482	50.0	4300	3.2158	1.0	15.7291	3.7429

Framework versions

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