akhisreelibra/malayalam-summariser

mt5-small-finetuned-oneindia

This model is a fine-tuned version of google/mt5-small on https://www.kaggle.com/datasets/akhisreelibra/malayalam-news dataset. It achieves the following results on the evaluation set:

• Loss: 2.0277
• Rouge1: 6.1146
• Rouge2: 0.9858
• Rougel: 6.085
• Rougelsum: 6.0965

Model description

Model outputs summarised one liner to an article.

Intended uses & limitations

Intended use in news domain, can be used to generate headlines to news articles.

Training and evaluation data

https://www.kaggle.com/datasets/akhisreelibra/malayalam-news

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5.6e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 8

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
3.1789	1.0	8994	2.2686	5.5492	0.8569	5.5448	5.5362
2.5859	2.0	17988	2.1751	5.7343	0.9775	5.7166	5.7195
2.4438	3.0	26982	2.1082	5.9718	0.9917	5.9503	5.9586
2.3584	4.0	35976	2.0730	6.1837	1.0286	6.1562	6.167
2.3015	5.0	44970	2.0512	6.0863	0.9656	6.0598	6.0643
2.2609	6.0	53964	2.0440	6.2119	0.9653	6.1859	6.1899
2.2344	7.0	62958	2.0303	6.2424	0.9835	6.2121	6.219
2.2166	8.0	71952	2.0277	6.1146	0.9858	6.085	6.0965

Framework versions

• Transformers 4.20.0
• Pytorch 1.11.0
• Datasets 2.1.0
• Tokenizers 0.12.1