velectra-base_v2

This model is a fine-tuned version of FPTAI/velectra-base-discriminator-cased on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5503
• Accuracy: 0.9242
• Precision Macro: 0.8370
• Recall Macro: 0.7946
• F1 Macro: 0.8125
• F1 Weighted: 0.9222

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: 3e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision Macro	Recall Macro	F1 Macro	F1 Weighted
0.5452	1.0	90	0.2734	0.9071	0.8647	0.6926	0.7190	0.8965
0.2546	2.0	180	0.2530	0.9198	0.8318	0.7882	0.8059	0.9176
0.1788	3.0	270	0.2528	0.9223	0.8241	0.7732	0.7929	0.9193
0.1323	4.0	360	0.2605	0.9261	0.8473	0.8000	0.8197	0.9241
0.0901	5.0	450	0.2840	0.9305	0.8839	0.7986	0.8303	0.9276
0.0682	6.0	540	0.3434	0.9210	0.8458	0.8007	0.8197	0.9192
0.0482	7.0	630	0.3689	0.9191	0.7970	0.8197	0.8073	0.9206
0.0443	8.0	720	0.3906	0.9223	0.8315	0.7728	0.7952	0.9191
0.0275	9.0	810	0.4178	0.9210	0.8717	0.7504	0.7861	0.9155
0.028	10.0	900	0.4642	0.9103	0.7837	0.7837	0.7835	0.9103
0.02	11.0	990	0.4823	0.9179	0.8459	0.7694	0.7971	0.9143
0.0122	12.0	1080	0.5070	0.9179	0.8594	0.7853	0.8136	0.9151
0.0098	13.0	1170	0.5093	0.9248	0.8387	0.7911	0.8106	0.9225
0.0108	14.0	1260	0.5309	0.9248	0.8678	0.7783	0.8098	0.9212
0.0101	15.0	1350	0.5214	0.9261	0.8623	0.7669	0.7986	0.9216
0.0076	16.0	1440	0.5352	0.9242	0.8653	0.7737	0.8054	0.9203
0.0042	17.0	1530	0.5533	0.9198	0.8163	0.7870	0.8000	0.9181
0.0058	18.0	1620	0.5503	0.9255	0.8574	0.7871	0.8138	0.9225
0.0034	19.0	1710	0.5590	0.9248	0.8349	0.8035	0.8173	0.9233
0.0029	20.0	1800	0.5503	0.9242	0.8370	0.7946	0.8125	0.9222

Framework versions

• Transformers 4.55.0
• Pytorch 2.7.0+cu126
• Datasets 4.0.0
• Tokenizers 0.21.4