summarizer-tamil-mbart

This model is a fine-tuned version of facebook/mbart-large-50-many-to-many-mmt on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.4252
• Rouge1: 9.3056
• Rouge2: 2.0
• Rougel: 9.2889
• Rougelsum: 9.2222
• Gen Len: 39.2233

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: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 5
• label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Gen Len	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
4.5487	0.2963	200	39.11	4.1027	2.1667	0.3333	2.1111	2.1667
4.0997	0.5926	400	41.7633	3.9744	2.2222	0.3333	2.2593	2.2222
4.0165	0.8889	600	52.0967	3.9417	2.6667	0.6667	2.6667	2.7778
3.7801	1.1852	800	45.3967	3.9424	2.1444	0.0	2.2222	2.1333
3.7308	1.4815	1000	41.3833	3.9573	2.7333	0.2222	2.6063	2.7905
3.7946	1.7778	1200	35.37	3.8979	1.0571	0.2222	0.9619	1.0571
3.6338	2.0741	1400	30.9567	3.9569	1.6611	0.3333	1.6333	1.6833
3.2282	2.3704	1600	42.4933	3.0726	4.0698	0.3889	3.9754	3.9825
3.1351	2.6667	1800	38.48	3.0771	2.8333	0.0	2.8095	2.8333
3.1739	2.9630	2000	40.04	3.0871	2.4921	0.0	2.496	2.4762
2.8247	3.2593	2200	39.95	3.0882	3.4706	0.2222	3.4421	3.4357
2.7748	3.5556	2400	38.29	3.0735	3.0	0.0	3.0	3.0
2.5244	3.8519	2600	2.4450	7.3889	1.2222	7.4667	7.5	38.1767
2.5382	4.1481	2800	2.4365	8.1111	1.9744	8.2111	8.1667	39.3333
2.4642	4.4444	3000	2.4334	8.3889	2.1905	8.5389	8.4444	37.7767
2.4641	4.7407	3200	2.4252	9.3056	2.0	9.2889	9.2222	39.2233

Framework versions

• Transformers 4.41.0
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1