5e2c020caea3631df1448f618c055624

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

• Loss: 2.0568
• Data Size: 1.0
• Epoch Runtime: 151.4874
• Bleu: 9.0567

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	16.0469	0	13.5759	0.2792
No log	1	1000	15.5384	0.0078	15.7546	0.2854
No log	2	2000	14.2046	0.0156	15.9420	0.3117
No log	3	3000	11.0596	0.0312	18.4288	0.3507
0.5046	4	4000	7.2043	0.0625	23.8459	0.3804
7.5837	5	5000	5.0983	0.125	31.3535	0.9019
0.354	6	6000	4.1686	0.25	48.7612	3.0741
0.4226	7	7000	3.5179	0.5	84.0710	2.5292
4.0056	8.0	8000	3.0932	1.0	154.1716	3.7938
3.7016	9.0	9000	2.9239	1.0	154.7242	4.4936
3.5164	10.0	10000	2.8166	1.0	154.9541	4.9045
3.3913	11.0	11000	2.7394	1.0	154.4091	5.2539
3.2694	12.0	12000	2.6767	1.0	154.1653	5.5500
3.1879	13.0	13000	2.6226	1.0	154.7862	5.8114
3.1059	14.0	14000	2.5770	1.0	152.5055	6.0134
3.0226	15.0	15000	2.5310	1.0	152.2572	6.2154
2.9793	16.0	16000	2.5003	1.0	152.6136	6.4185
2.8864	17.0	17000	2.4742	1.0	152.7497	6.5705
2.8905	18.0	18000	2.4377	1.0	153.1263	6.7235
2.8353	19.0	19000	2.4112	1.0	153.0815	6.8518
2.7771	20.0	20000	2.3906	1.0	152.9901	7.0136
2.7758	21.0	21000	2.3625	1.0	153.3342	7.1366
2.7155	22.0	22000	2.3436	1.0	151.8810	7.2288
2.6817	23.0	23000	2.3251	1.0	151.9600	7.3268
2.636	24.0	24000	2.3105	1.0	153.9415	7.4365
2.6669	25.0	25000	2.2921	1.0	153.0338	7.5221
2.5816	26.0	26000	2.2731	1.0	151.8625	7.6526
2.5615	27.0	27000	2.2599	1.0	151.9675	7.7261
2.4947	28.0	28000	2.2530	1.0	151.5303	7.8470
2.5166	29.0	29000	2.2336	1.0	153.0677	7.9289
2.4557	30.0	30000	2.2256	1.0	152.1012	8.0204
2.474	31.0	31000	2.2065	1.0	154.6577	8.0815
2.4297	32.0	32000	2.1982	1.0	153.2791	8.1406
2.3823	33.0	33000	2.1887	1.0	151.7204	8.2139
2.3719	34.0	34000	2.1878	1.0	153.3640	8.2638
2.3598	35.0	35000	2.1707	1.0	152.8024	8.3561
2.3231	36.0	36000	2.1591	1.0	154.2330	8.4129
2.3012	37.0	37000	2.1555	1.0	153.5931	8.4714
2.317	38.0	38000	2.1326	1.0	153.1331	8.5374
2.2725	39.0	39000	2.1347	1.0	152.3129	8.5907
2.2476	40.0	40000	2.1373	1.0	156.9343	8.6044
2.2446	41.0	41000	2.1207	1.0	156.4713	8.6799
2.2218	42.0	42000	2.1083	1.0	152.9422	8.7118
2.2042	43.0	43000	2.1066	1.0	154.8949	8.7772
2.2087	44.0	44000	2.1007	1.0	153.7430	8.7890
2.1733	45.0	45000	2.0951	1.0	153.8691	8.8444
2.193	46.0	46000	2.0931	1.0	151.8176	8.8520
2.1578	47.0	47000	2.0772	1.0	152.4861	8.9144
2.1741	48.0	48000	2.0803	1.0	152.9961	8.9398
2.1133	49.0	49000	2.0752	1.0	152.8284	9.0189
2.0837	50.0	50000	2.0568	1.0	151.4874	9.0567

Framework versions

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