gpt2-gpt2-mc-weight0.25-epoch15-new

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

• Loss: 4.7276
• Cls loss: 3.0579
• Lm loss: 3.9626
• Cls Accuracy: 0.6110
• Cls F1: 0.6054
• Cls Precision: 0.6054
• Cls Recall: 0.6110
• Perplexity: 52.59

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: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Cls loss	Lm loss	Cls Accuracy	Cls F1	Cls Precision	Cls Recall	Perplexity
4.674	1.0	3470	4.4372	1.5961	4.0380	0.5487	0.5279	0.5643	0.5487	56.71
4.3809	2.0	6940	4.3629	1.4483	4.0006	0.6023	0.5950	0.6174	0.6023	54.63
4.2522	3.0	10410	4.3721	1.5476	3.9849	0.6012	0.5981	0.6186	0.6012	53.78
4.1478	4.0	13880	4.3892	1.6429	3.9782	0.6081	0.6019	0.6128	0.6081	53.42
4.0491	5.0	17350	4.4182	1.8093	3.9656	0.6156	0.6091	0.6163	0.6156	52.75
3.9624	6.0	20820	4.4757	2.0348	3.9666	0.6121	0.6048	0.6189	0.6121	52.81
3.8954	7.0	24290	4.4969	2.1327	3.9634	0.6092	0.6028	0.6087	0.6092	52.64
3.846	8.0	27760	4.5632	2.4063	3.9613	0.6017	0.5972	0.6014	0.6017	52.52
3.8036	9.0	31230	4.6068	2.5888	3.9592	0.6052	0.5988	0.6026	0.6052	52.41
3.7724	10.0	34700	4.6175	2.6197	3.9621	0.6052	0.6006	0.6009	0.6052	52.57
3.7484	11.0	38170	4.6745	2.8470	3.9622	0.6046	0.5996	0.6034	0.6046	52.57
3.7291	12.0	41640	4.6854	2.8950	3.9611	0.6110	0.6056	0.6049	0.6110	52.52
3.7148	13.0	45110	4.7103	2.9919	3.9618	0.6063	0.6002	0.6029	0.6063	52.55
3.703	14.0	48580	4.7226	3.0417	3.9616	0.6081	0.6027	0.6021	0.6081	52.54
3.6968	15.0	52050	4.7276	3.0579	3.9626	0.6110	0.6054	0.6054	0.6110	52.59

Framework versions

• Transformers 4.21.2
• Pytorch 1.12.1
• Datasets 2.4.0
• Tokenizers 0.12.1