breadlicker45/multilingual-gender-v2

Model Description

This is a model for classifying male, female, and non-binary genders from one paragraph.

Training Details

• batch-size: 32
• epochs: 3
• GPU used: An Nvidia P40 gpu

Evaluation

• eval_loss: 0.652
• eval_f1: 0.593
• eval_roc_auc: 0.616
• eval_accuracy: 0.30039
• eval_runtime: 39.7364
• eval_samples_per_second: 308.961
• eval_steps_per_second: 9.664
• epoch: 3.0

training Metrics

Step	Training Loss	Validation Loss	F1	ROC AUC	Accuracy
500	0.675000	0.666651	0.5384	0.593959	0.303494
1000	0.667200	0.665188	0.5496	0.597031	0.302924
1500	0.662900	0.661620	0.5836	0.602752	0.287204
2000	0.659800	0.662705	0.5710	0.602509	0.289240
2500	0.660200	0.664614	0.5511	0.600784	0.303902
3000	0.659100	0.658421	0.5650	0.604483	0.304716
3500	0.650300	0.657569	0.5821	0.609502	0.300236
4000	0.648100	0.654424	0.5830	0.609785	0.293720
4500	0.640700	0.654051	0.5743	0.612857	0.308544
5000	0.645500	0.651678	0.5806	0.613973	0.305531
5500	0.642400	0.651911	0.5808	0.614797	0.307893
6000	0.642100	0.651853	0.5795	0.616014	0.312861
6500	0.628700	0.653005	0.5909	0.616887	0.304472
7000	0.624900	0.653188	0.5849	0.616239	0.306264
7500	0.623600	0.652131	0.5938	0.616488	0.300399
8000	0.622500	0.652739	0.5855	0.617415	0.310418
8500	0.622300	0.651849	0.5916	0.617908	0.308056
9000	0.622400	0.651472	0.5910	0.618263	0.308137