electra-large-discriminator-nli-efl-hateval

This model is a fine-tuned version of ynie/electra-large-discriminator-snli_mnli_fever_anli_R1_R2_R3-nli on the None dataset. It achieves the following results on the evaluation set:

• Accuracy: 0.798
• F1: 0.7968
• Loss: 0.4166

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-06
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Accuracy	F1	Validation Loss
0.4175	1.0	210	0.7317	0.7305	0.4020
0.3061	2.0	420	0.768	0.7675	0.3520
0.2588	3.0	630	0.79	0.7888	0.3253
0.234	4.0	840	0.788	0.7877	0.3373
0.2116	5.0	1050	0.804	0.8033	0.3247
0.1974	6.0	1260	0.793	0.7928	0.3400
0.1807	7.0	1470	0.7973	0.7969	0.3511
0.1715	8.0	1680	0.7993	0.7989	0.3496
0.1577	9.0	1890	0.8043	0.8032	0.3507
0.1469	10.0	2100	0.798	0.7970	0.3604
0.1394	11.0	2310	0.7967	0.7957	0.3734
0.1322	12.0	2520	0.7913	0.7906	0.3929
0.1231	13.0	2730	0.795	0.7941	0.3954
0.1189	14.0	2940	0.7977	0.7963	0.3994
0.1143	15.0	3150	0.7993	0.7980	0.3995
0.1083	16.0	3360	0.7927	0.7918	0.4125
0.1079	17.0	3570	0.7993	0.7979	0.4036
0.1055	18.0	3780	0.7967	0.7956	0.4121
0.1006	19.0	3990	0.7973	0.7961	0.4152
0.101	20.0	4200	0.798	0.7968	0.4166

Framework versions

• Transformers 4.17.0
• Pytorch 1.11.0+cu113
• Datasets 2.0.0
• Tokenizers 0.11.6