w2v-bert-2.0-CV_Fleurs-lg-100hrs-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: 0.2835
• Wer: 0.2749
• Cer: 0.0546

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
1.3102	1.0	7057	0.3038	0.3739	0.0787
0.1944	2.0	14114	0.2340	0.3188	0.0634
0.1558	3.0	21171	0.2157	0.3021	0.0598
0.1373	4.0	28228	0.2037	0.2926	0.0574
0.1237	5.0	35285	0.2079	0.3025	0.0591
0.1152	6.0	42342	0.2044	0.2899	0.0583
0.1091	7.0	49399	0.2025	0.3012	0.0567
0.1042	8.0	56456	0.2033	0.2728	0.0552
0.0985	9.0	63513	0.2035	0.2760	0.0543
0.0949	10.0	70570	0.2054	0.2730	0.0541
0.0885	11.0	77627	0.2071	0.2758	0.0557
0.08	12.0	84684	0.2039	0.2743	0.0540
0.0719	13.0	91741	0.2045	0.2648	0.0531
0.063	14.0	98798	0.2133	0.2682	0.0544
0.0555	15.0	105855	0.2107	0.2658	0.0524
0.0471	16.0	112912	0.2408	0.2687	0.0529
0.0399	17.0	119969	0.2419	0.2721	0.0535
0.0341	18.0	127026	0.2587	0.2833	0.0551
0.0293	19.0	134083	0.2518	0.2735	0.0534
0.0255	20.0	141140	0.2795	0.2669	0.0530
0.022	21.0	148197	0.2748	0.2820	0.0543
0.0196	22.0	155254	0.2837	0.2785	0.0529
0.017	23.0	162311	0.2835	0.2749	0.0546

Framework versions

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