medium

This model is a fine-tuned version of prithivida/parrot_paraphraser_on_T5 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6025
• Rouge1: 81.6007
• Rouge2: 75.1196
• Rougel: 81.4213
• Rougelsum: 81.4956
• Gen Len: 32.4286

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: 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: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
No log	1.0	63	0.5775	65.0748	58.8985	64.5731	63.6249	19.0
No log	2.0	126	0.5806	74.3055	69.2025	73.4922	73.0941	17.8571
No log	3.0	189	0.6025	71.3808	66.0359	70.1235	69.4614	18.0

Framework versions

• Transformers 4.15.0
• Pytorch 1.10.1+cu113
• Datasets 1.17.0
• Tokenizers 0.10.3