english-marathi-colloquial-translator1

This model is a fine-tuned version of Helsinki-NLP/opus-mt-en-mr on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5252

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: 0.0003
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 2
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
15.4233	0.3636	2	7.9121
13.0588	0.7273	4	7.9121
8.4646	1.0	6	7.9121
4.474	1.3636	8	3.3942
1.7662	1.7273	10	0.6550
0.7863	2.0	12	0.5941
1.1458	2.3636	14	0.5240
0.8619	2.7273	16	0.4977
0.7108	3.0	18	0.4929
0.8382	3.3636	20	0.4984
0.5416	3.7273	22	0.4854
0.2726	4.0	24	0.4787
0.4561	4.3636	26	0.4806
0.5203	4.7273	28	0.4890
0.233	5.0	30	0.4989
0.4509	5.3636	32	0.4978
0.3384	5.7273	34	0.4999
0.1781	6.0	36	0.5054
0.3225	6.3636	38	0.5101
0.2737	6.7273	40	0.5159
0.0695	7.0	42	0.5239
0.2417	7.3636	44	0.5263
0.3237	7.7273	46	0.5253
0.1458	8.0	48	0.5249
0.2153	8.3636	50	0.5252

Framework versions

• Transformers 4.47.1
• Pytorch 2.6.0+cu124
• Datasets 3.3.1
• Tokenizers 0.21.0