Helsinki_sg_inf_en

This model is a fine-tuned version of MubarakB/Helsinki_lg_inf_en on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2858
• Bleu: 5.572
• Gen Len: 9.2338

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
No log	1.0	173	0.6201	0.0669	6.4272
No log	2.0	346	0.5838	0.0805	5.9299
0.6387	3.0	519	0.5608	0.0808	6.1499
0.6387	4.0	692	0.5425	0.1017	8.1706
0.6387	5.0	865	0.5253	0.1307	8.0857
0.5575	6.0	1038	0.5073	0.1321	9.5706
0.5575	7.0	1211	0.4894	0.1799	12.0178
0.5575	8.0	1384	0.4735	0.4182	14.6461
0.5174	9.0	1557	0.4570	0.6454	8.5702
0.5174	10.0	1730	0.4399	0.7703	10.7481
0.5174	11.0	1903	0.4245	1.0412	10.367
0.4806	12.0	2076	0.4091	0.9907	11.4105
0.4806	13.0	2249	0.3942	1.3776	10.6073
0.4806	14.0	2422	0.3828	1.682	9.6396
0.4459	15.0	2595	0.3686	1.9324	10.269
0.4459	16.0	2768	0.3588	2.1982	10.6911
0.4459	17.0	2941	0.3477	2.7115	10.718
0.4168	18.0	3114	0.3381	3.0707	9.9858
0.4168	19.0	3287	0.3296	3.327	9.5866
0.4168	20.0	3460	0.3212	3.6662	9.3844
0.3955	21.0	3633	0.3153	4.0484	9.2152
0.3955	22.0	3806	0.3087	4.5079	9.3263
0.3955	23.0	3979	0.3026	4.7791	9.8225
0.3756	24.0	4152	0.2987	5.0946	9.6704
0.3756	25.0	4325	0.2950	5.1408	9.2835
0.3756	26.0	4498	0.2917	5.3052	9.277
0.3633	27.0	4671	0.2891	5.4256	9.7289
0.3633	28.0	4844	0.2873	5.5373	9.5742
0.3563	29.0	5017	0.2861	5.5454	10.1187
0.3563	30.0	5190	0.2858	5.572	9.2338

Framework versions

• Transformers 4.45.1
• Pytorch 2.4.0
• Datasets 3.0.1
• Tokenizers 0.20.0