m2m100_418M-finetuned-en-to-hi

This model is a fine-tuned version of facebook/m2m100_418M on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.0453
• Bleu: 17.4993
• Gen Len: 6.7284

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: 48
• eval_batch_size: 48
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
2.4274	0.16	500	2.1152	4.4935	6.8813
2.1915	0.33	1000	1.9722	5.8486	6.9727
2.1187	0.49	1500	1.8575	5.5802	6.9993
2.0151	0.66	2000	1.7686	8.8892	6.8233
1.9709	0.82	2500	1.6948	8.4082	6.8809
1.9376	0.99	3000	1.6341	10.0801	6.85
1.761	1.15	3500	1.5788	8.1916	6.8816
1.7269	1.32	4000	1.5380	10.2779	6.9447
1.7231	1.48	4500	1.4946	6.9244	6.9402
1.6925	1.65	5000	1.4456	13.7246	6.9018
1.6658	1.81	5500	1.4146	9.1181	6.9104
1.6673	1.98	6000	1.3727	8.6535	6.8682
1.5165	2.14	6500	1.3441	14.8146	6.9804
1.5111	2.31	7000	1.3101	11.192	6.92
1.4889	2.47	7500	1.2814	11.8364	6.9509
1.4903	2.64	8000	1.2510	16.8035	6.9316
1.4871	2.8	8500	1.2298	14.5766	6.9053
1.4854	2.97	9000	1.2051	14.2822	6.8438
1.3719	3.13	9500	1.1758	16.1779	6.8918
1.3481	3.3	10000	1.1612	20.1789	6.8138
1.3585	3.46	10500	1.1410	15.6937	6.8613
1.35	3.63	11000	1.1261	20.0808	6.832
1.3557	3.79	11500	1.1069	19.588	6.8242
1.3329	3.96	12000	1.0924	19.9913	6.796
1.2792	4.12	12500	1.0791	18.8275	6.7616
1.2568	4.29	13000	1.0701	16.7189	6.7676
1.2558	4.45	13500	1.0605	18.7687	6.7464
1.2533	4.62	14000	1.0541	19.1818	6.7693
1.2559	4.78	14500	1.0475	19.0462	6.738
1.2513	4.95	15000	1.0453	17.4993	6.7284

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.2+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0