english-marathi-colloquial-translator

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: 0.4788

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
14.2862	0.1818	2	8.4218
14.6418	0.3636	4	8.4218
14.3291	0.5455	6	8.4218
4.6555	0.7273	8	3.0796
1.5096	0.9091	10	0.6436
1.7952	1.0909	12	0.5874
1.0955	1.2727	14	0.4750
1.1103	1.4545	16	0.4446
0.7535	1.6364	18	0.4273
0.726	1.8182	20	0.4126
0.7479	2.0	22	0.4048
0.4539	2.1818	24	0.4023
0.5944	2.3636	26	0.4068
0.703	2.5455	28	0.4104
0.5892	2.7273	30	0.4079
0.483	2.9091	32	0.4073
0.5133	3.0909	34	0.4119
0.4196	3.2727	36	0.4136
0.3731	3.4545	38	0.4158
0.4221	3.6364	40	0.4191
0.6552	3.8182	42	0.4218
0.3614	4.0	44	0.4225
0.3733	4.1818	46	0.4252
0.2367	4.3636	48	0.4319
0.4236	4.5455	50	0.4405
0.4277	4.7273	52	0.4408
0.3173	4.9091	54	0.4393
0.172	5.0909	56	0.4421
0.2636	5.2727	58	0.4468
0.2434	5.4545	60	0.4483
0.2083	5.6364	62	0.4488
0.3047	5.8182	64	0.4472
0.1746	6.0	66	0.4515
0.1768	6.1818	68	0.4566
0.1424	6.3636	70	0.4596
0.142	6.5455	72	0.4618
0.1157	6.7273	74	0.4671
0.1909	6.9091	76	0.4706
0.1102	7.0909	78	0.4708
0.2191	7.2727	80	0.4724
0.1248	7.4545	82	0.4739
0.2527	7.6364	84	0.4758
0.1135	7.8182	86	0.4789
0.1232	8.0	88	0.4808
0.1503	8.1818	90	0.4814
0.1294	8.3636	92	0.4801
0.1287	8.5455	94	0.4788
0.1029	8.7273	96	0.4785
0.0962	8.9091	98	0.4778
0.103	9.0909	100	0.4775
0.0942	9.2727	102	0.4777
0.0695	9.4545	104	0.4781
0.0857	9.6364	106	0.4785
0.0787	9.8182	108	0.4788
0.1226	10.0	110	0.4788

Framework versions

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