04357355773e8961693280f56a1449a0

This model is a fine-tuned version of google/mt5-small on the Helsinki-NLP/opus_books [fr-it] dataset. It achieves the following results on the evaluation set:

• Loss: 2.3710
• Data Size: 1.0
• Epoch Runtime: 54.0672
• Bleu: 4.7280

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	25.1154	0	5.3825	0.0016
No log	1	367	23.2656	0.0078	6.6290	0.0018
No log	2	734	20.2779	0.0156	6.4812	0.0032
No log	3	1101	16.9192	0.0312	7.7052	0.0029
No log	4	1468	12.3286	0.0625	8.8945	0.0057
0.9075	5	1835	8.1736	0.125	11.9506	0.0114
8.2208	6	2202	4.8777	0.25	18.2376	0.0163
4.8968	7	2569	3.4010	0.5	30.3690	0.4978
4.1438	8.0	2936	3.1225	1.0	55.0431	1.2713
3.8377	9.0	3303	2.9898	1.0	55.7866	1.6002
3.6606	10.0	3670	2.9150	1.0	55.9007	1.8640
3.5493	11.0	4037	2.8568	1.0	56.5953	2.1198
3.4442	12.0	4404	2.8096	1.0	55.7102	2.2761
3.403	13.0	4771	2.7733	1.0	54.4628	2.4571
3.3022	14.0	5138	2.7394	1.0	55.6655	2.5954
3.2628	15.0	5505	2.7122	1.0	54.6192	2.7575
3.1883	16.0	5872	2.6843	1.0	55.1374	2.8996
3.1465	17.0	6239	2.6687	1.0	54.9305	2.9907
3.1012	18.0	6606	2.6433	1.0	55.3830	3.0827
3.0873	19.0	6973	2.6317	1.0	56.0201	3.1600
3.0363	20.0	7340	2.6021	1.0	54.8430	3.2487
2.9852	21.0	7707	2.5877	1.0	55.4414	3.3326
2.9617	22.0	8074	2.5818	1.0	55.4208	3.4285
2.9161	23.0	8441	2.5707	1.0	54.6723	3.4842
2.8687	24.0	8808	2.5546	1.0	54.5485	3.5617
2.8744	25.0	9175	2.5502	1.0	54.6917	3.6545
2.8176	26.0	9542	2.5291	1.0	54.6620	3.7157
2.7957	27.0	9909	2.5210	1.0	54.5980	3.8414
2.7676	28.0	10276	2.5061	1.0	56.9414	3.8788
2.7649	29.0	10643	2.5044	1.0	56.3715	3.9675
2.743	30.0	11010	2.4944	1.0	55.0688	4.0330
2.697	31.0	11377	2.4803	1.0	54.9879	4.1151
2.7005	32.0	11744	2.4716	1.0	54.7681	4.1206
2.6596	33.0	12111	2.4614	1.0	54.9092	4.2020
2.6309	34.0	12478	2.4564	1.0	56.2777	4.2420
2.6054	35.0	12845	2.4494	1.0	55.9621	4.2832
2.6204	36.0	13212	2.4498	1.0	55.7327	4.3257
2.586	37.0	13579	2.4346	1.0	55.3127	4.3834
2.583	38.0	13946	2.4379	1.0	55.0265	4.4316
2.5634	39.0	14313	2.4232	1.0	55.3696	4.4535
2.5348	40.0	14680	2.4112	1.0	54.8842	4.5320
2.5346	41.0	15047	2.4179	1.0	56.2262	4.5211
2.5066	42.0	15414	2.4176	1.0	55.7012	4.6158
2.4928	43.0	15781	2.4052	1.0	55.0442	4.6144
2.4923	44.0	16148	2.4027	1.0	54.6755	4.6431
2.4631	45.0	16515	2.3979	1.0	54.2127	4.6744
2.4756	46.0	16882	2.3845	1.0	57.4652	4.6653
2.4326	47.0	17249	2.3916	1.0	54.8645	4.6903
2.4329	48.0	17616	2.3900	1.0	54.0670	4.6948
2.3847	49.0	17983	2.3755	1.0	53.8656	4.7380
2.4299	50.0	18350	2.3710	1.0	54.0672	4.7280

Framework versions

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