jialicheng/unlearn_samsum_t5-base_salun_4_42

samsum_42

This model is a fine-tuned version of google/t5-v1_1-base on the samsum dataset. It achieves the following results on the evaluation set:

• Loss: 1.4769
• Rouge1: 49.78
• Rouge2: 25.0211
• Rougel: 41.3044
• Rougelsum: 45.928
• Gen Len: 24.1523
• Test Rougel: 41.2383
• Df Rougel: 41.8167
• Unlearn Overall Rougel: 0.2108
• Unlearn Time: 4034.7489

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: 32
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len	Overall Rougel	Unlearn Overall Rougel	Time
No log	1.0	460	1.4973	48.6265	24.2942	41.7917	44.7761	22.6510	-0.0701	-0.0701	-1
No log	2.0	920	1.4938	48.9671	24.4584	42.125	45.2967	24.3021	-0.2166	-0.2166	-1
1.886	3.0	1380	1.4769	49.78	25.0211	41.8167	45.928	24.1523	0.2108	0.2108	-1
1.886	4.0	1840	1.4737	49.4167	24.7497	41.9533	45.6082	23.8464	0.0736	0.0736	-1
1.7917	5.0	2300	1.4701	49.5761	24.7762	42.1164	45.7806	23.8906	0.0526	0.0526	-1

Framework versions

• Transformers 4.39.3
• Pytorch 2.3.0+cu121
• Datasets 2.19.0
• Tokenizers 0.15.2