f5b7419600bbc19f534df6daa8c13887

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

• Loss: 3.1133
• Data Size: 1.0
• Epoch Runtime: 9.9173
• Bleu: 1.4951

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	23.3744	0	1.6147	0.0010
No log	1	58	23.1855	0.0078	1.6773	0.0015
No log	2	116	22.8682	0.0156	1.7411	0.0015
No log	3	174	22.4285	0.0312	2.1981	0.0013
No log	4	232	21.2909	0.0625	2.5757	0.0015
No log	5	290	18.6704	0.125	2.9737	0.0014
1.9421	6	348	15.5708	0.25	3.9325	0.0021
2.5078	7	406	10.8684	0.5	5.6988	0.0028
8.8154	8.0	464	6.1040	1.0	9.7775	0.0222
7.3386	9.0	522	4.9965	1.0	9.5529	0.0366
5.9974	10.0	580	4.2617	1.0	9.8360	0.0584
5.2919	11.0	638	3.9682	1.0	9.5438	0.0889
4.9133	12.0	696	3.8141	1.0	9.8508	0.4793
4.5573	13.0	754	3.7247	1.0	9.5140	0.5970
4.4588	14.0	812	3.6695	1.0	9.4895	0.6336
4.3457	15.0	870	3.6156	1.0	9.5015	0.6482
4.2786	16.0	928	3.5704	1.0	9.8842	0.7365
4.2259	17.0	986	3.5328	1.0	10.2678	0.8037
4.1609	18.0	1044	3.5001	1.0	10.2281	0.8533
4.078	19.0	1102	3.4749	1.0	10.2322	0.8296
4.0242	20.0	1160	3.4503	1.0	10.8015	0.9442
3.9846	21.0	1218	3.4242	1.0	9.7304	0.9094
3.9568	22.0	1276	3.4005	1.0	9.6782	0.9428
3.923	23.0	1334	3.3833	1.0	9.8428	0.9771
3.8988	24.0	1392	3.3621	1.0	9.8737	1.0161
3.8403	25.0	1450	3.3472	1.0	9.9675	1.0115
3.8154	26.0	1508	3.3342	1.0	10.4292	1.0457
3.7767	27.0	1566	3.3146	1.0	10.8942	1.0446
3.7655	28.0	1624	3.3019	1.0	10.8581	1.0148
3.7476	29.0	1682	3.2869	1.0	10.8080	1.1099
3.7219	30.0	1740	3.2732	1.0	9.6816	1.1335
3.6954	31.0	1798	3.2618	1.0	10.1066	1.1724
3.6559	32.0	1856	3.2502	1.0	10.3752	1.1747
3.6351	33.0	1914	3.2424	1.0	10.3499	1.1920
3.622	34.0	1972	3.2301	1.0	10.6199	1.2463
3.5962	35.0	2030	3.2204	1.0	10.7700	1.2873
3.5807	36.0	2088	3.2108	1.0	10.9988	1.2177
3.5699	37.0	2146	3.2041	1.0	11.6971	1.2554
3.5291	38.0	2204	3.1918	1.0	11.6496	1.2671
3.5063	39.0	2262	3.1890	1.0	11.9242	1.2835
3.5058	40.0	2320	3.1768	1.0	9.8241	1.3632
3.4813	41.0	2378	3.1714	1.0	10.3065	1.3329
3.4609	42.0	2436	3.1621	1.0	10.6125	1.3574
3.454	43.0	2494	3.1562	1.0	10.9934	1.3835
3.4304	44.0	2552	3.1523	1.0	11.1347	1.3835
3.4036	45.0	2610	3.1419	1.0	11.0001	1.4078
3.4139	46.0	2668	3.1367	1.0	11.2890	1.5005
3.3835	47.0	2726	3.1305	1.0	11.2280	1.4660
3.3731	48.0	2784	3.1234	1.0	12.6309	1.4660
3.3645	49.0	2842	3.1187	1.0	11.7803	1.5160
3.3363	50.0	2900	3.1133	1.0	9.9173	1.4951

Framework versions

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