FLANG-ELECTRA_Synonym-wordnet

This model is a fine-tuned version of SALT-NLP/FLANG-ELECTRA on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3701
• Accuracy: 0.9376
• F1: 0.9374
• Precision: 0.9374
• Recall: 0.9376

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: 0.0001
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
0.6271	1.0	181	0.4625	0.8409	0.8408	0.8424	0.8409
0.3372	2.0	362	0.3312	0.8955	0.8959	0.8991	0.8955
0.1762	3.0	543	0.3046	0.9048	0.9040	0.9050	0.9048
0.313	4.0	724	0.3908	0.8986	0.8988	0.9018	0.8986
0.2564	5.0	905	0.3268	0.9080	0.9073	0.9078	0.9080
0.3189	6.0	1086	0.6418	0.7956	0.7933	0.8130	0.7956
0.365	7.0	1267	0.7276	0.7894	0.7889	0.7890	0.7894
0.1356	8.0	1448	0.6091	0.8814	0.8811	0.8816	0.8814
0.1139	9.0	1629	0.4184	0.8924	0.8916	0.8948	0.8924
0.1238	10.0	1810	0.3155	0.9220	0.9213	0.9231	0.9220
0.077	11.0	1991	0.4511	0.9017	0.9025	0.9061	0.9017
0.0613	12.0	2172	0.4132	0.9142	0.9141	0.9145	0.9142
0.0514	13.0	2353	0.3735	0.9298	0.9296	0.9321	0.9298
0.057	14.0	2534	0.3701	0.9376	0.9374	0.9374	0.9376
0.0152	15.0	2715	0.3872	0.9360	0.9357	0.9384	0.9360
0.0236	16.0	2896	0.4117	0.9314	0.9310	0.9320	0.9314
0.0277	17.0	3077	0.5325	0.9204	0.9197	0.9208	0.9204
0.0021	18.0	3258	0.4227	0.9236	0.9229	0.9236	0.9236
0.0005	19.0	3439	0.5409	0.9314	0.9308	0.9334	0.9314

Framework versions

• Transformers 4.37.0
• Pytorch 2.1.2
• Datasets 2.1.0
• Tokenizers 0.15.1