ucla-nb-project/bart-finetuned

bart-base-finetuned-xe_ey_fae

This model is a fine-tuned version of facebook/bart-base on the datasets/all_binary_and_xe_ey_fae_counterfactual dataset. It achieves the following results on the evaluation set:

• Loss: 1.3945
• Accuracy: 0.7180

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: 1e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 100
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
5.4226	0.06	500	3.8138	0.3628
4.0408	0.12	1000	3.0576	0.4630
3.4979	0.18	1500	2.7016	0.5133
3.1691	0.24	2000	2.4880	0.5431
2.9564	0.3	2500	2.3309	0.5644
2.8078	0.35	3000	2.2320	0.5792
2.6741	0.41	3500	2.1506	0.5924
2.5323	0.47	4000	1.9846	0.6176
2.3678	0.53	4500	1.8813	0.6375
2.25	0.59	5000	1.8100	0.6497
2.1795	0.65	5500	1.7632	0.6579
2.1203	0.71	6000	1.7238	0.6646
2.0764	0.77	6500	1.6856	0.6713
2.026	0.83	7000	1.6569	0.6760
1.9942	0.89	7500	1.6309	0.6803
1.9665	0.95	8000	1.6122	0.6836
1.9395	1.0	8500	1.5913	0.6866
1.9155	1.06	9000	1.5758	0.6895
1.8828	1.12	9500	1.5607	0.6918
1.8721	1.18	10000	1.5422	0.6948
1.8474	1.24	10500	1.5320	0.6964
1.8293	1.3	11000	1.5214	0.6978
1.8129	1.36	11500	1.5102	0.6998
1.8148	1.42	12000	1.5010	0.7013
1.7903	1.48	12500	1.4844	0.7038
1.7815	1.54	13000	1.4823	0.7039
1.7637	1.6	13500	1.4746	0.7052
1.7623	1.66	14000	1.4701	0.7061
1.7402	1.71	14500	1.4598	0.7076
1.7376	1.77	15000	1.4519	0.7090
1.7287	1.83	15500	1.4501	0.7101
1.7273	1.89	16000	1.4409	0.7107
1.7119	1.95	16500	1.4314	0.7125
1.7098	2.01	17000	1.4269	0.7129
1.6978	2.07	17500	1.4275	0.7132
1.698	2.13	18000	1.4218	0.7140
1.6837	2.19	18500	1.4151	0.7147
1.6908	2.25	19000	1.4137	0.7149
1.6902	2.31	19500	1.4085	0.7161
1.6741	2.36	20000	1.4121	0.7154
1.6823	2.42	20500	1.4037	0.7165
1.6692	2.48	21000	1.4039	0.7164
1.6669	2.54	21500	1.4015	0.7172
1.6613	2.6	22000	1.3979	0.7179
1.664	2.66	22500	1.3960	0.7180
1.6615	2.72	23000	1.4012	0.7172
1.6627	2.78	23500	1.3974	0.7178
1.6489	2.84	24000	1.3948	0.7182
1.6429	2.9	24500	1.3921	0.7184
1.6477	2.96	25000	1.3910	0.7182

Framework versions

• Transformers 4.36.2
• Pytorch 2.2.0+cu121
• Datasets 2.17.0
• Tokenizers 0.15.2