w2v-bert-2.0-CV_Fleurs-lg-50hrs-v4

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3482
• Wer: 0.2832
• Cer: 0.0557

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: 8
• eval_batch_size: 4
• seed: 42
• 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
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.9126	1.0	3160	0.3415	0.4010	0.0853
0.2463	2.0	6320	0.2633	0.3447	0.0670
0.1946	3.0	9480	0.2369	0.3201	0.0633
0.168	4.0	12640	0.2246	0.3098	0.0607
0.15	5.0	15800	0.2179	0.3205	0.0595
0.1394	6.0	18960	0.2245	0.3060	0.0594
0.1283	7.0	22120	0.2173	0.3029	0.0600
0.1219	8.0	25280	0.2203	0.3183	0.0583
0.1155	9.0	28440	0.2148	0.2923	0.0573
0.1117	10.0	31600	0.2334	0.3037	0.0586
0.1031	11.0	34760	0.2162	0.2876	0.0578
0.0908	12.0	37920	0.2210	0.2883	0.0560
0.0804	13.0	41080	0.2271	0.3001	0.0581
0.0706	14.0	44240	0.2403	0.2753	0.0540
0.0602	15.0	47400	0.2528	0.2955	0.0578
0.0512	16.0	50560	0.2695	0.2883	0.0555
0.0432	17.0	53720	0.2597	0.2903	0.0554
0.0367	18.0	56880	0.2764	0.2850	0.0556
0.0317	19.0	60040	0.2954	0.2908	0.0570
0.0267	20.0	63200	0.3053	0.2878	0.0556
0.0236	21.0	66360	0.3087	0.2868	0.0565
0.0208	22.0	69520	0.2907	0.2970	0.0584
0.0175	23.0	72680	0.3274	0.2838	0.0550
0.0169	24.0	75840	0.3482	0.2832	0.0557

Framework versions

• Transformers 4.46.0
• Pytorch 2.1.0+cu118
• Datasets 3.0.2
• Tokenizers 0.20.1