taylor-swift-model-paragraphs

This model is a fine-tuned version of distilgpt2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 3.3564

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: 2e-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: 25

Training results

Training Loss	Epoch	Step	Validation Loss
3.9316	1.0	59	3.8227
3.824	2.0	118	3.7301
3.5808	3.0	177	3.6658
3.625	4.0	236	3.6205
3.643	5.0	295	3.5862
3.5443	6.0	354	3.5545
3.4535	7.0	413	3.5274
3.398	8.0	472	3.5072
3.3253	9.0	531	3.4833
3.4111	10.0	590	3.4688
3.3461	11.0	649	3.4503
3.3133	12.0	708	3.4373
3.3921	13.0	767	3.4246
3.2661	14.0	826	3.4102
3.2257	15.0	885	3.4052
3.1837	16.0	944	3.3911
3.1935	17.0	1003	3.3849
2.9369	18.0	1062	3.3774
3.2486	19.0	1121	3.3721
3.1542	20.0	1180	3.3681
3.0771	21.0	1239	3.3624
3.1206	22.0	1298	3.3581
3.0358	23.0	1357	3.3585
2.9207	24.0	1416	3.3568
3.0496	25.0	1475	3.3564

Framework versions

• Transformers 4.25.1
• Pytorch 1.13.0+cu116
• Datasets 2.8.0
• Tokenizers 0.13.2