wav2vec2-lv-60-espeak-cv-ft-custom_vocab-OtherDiacritics-ds-f3

This model is a fine-tuned version of facebook/wav2vec2-lv-60-espeak-cv-ft on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2725
• Per: 0.2806

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: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Per
17.7086	0.7194	400	4.5721	0.9888
4.3976	1.4388	800	4.0921	0.9888
4.0518	2.1583	1200	3.5720	0.9888
3.1677	2.8777	1600	1.4785	0.4295
2.0338	3.5971	2000	0.8687	0.3483
1.5881	4.3165	2400	0.6268	0.3161
1.3712	5.0360	2800	0.5133	0.3033
1.2124	5.7554	3200	0.4322	0.2952
1.126	6.4748	3600	0.3897	0.2912
1.0368	7.1942	4000	0.3668	0.2898
0.9932	7.9137	4400	0.3369	0.2865
0.9417	8.6331	4800	0.3338	0.2866
0.8739	9.3525	5200	0.3261	0.2848
0.8668	10.0719	5600	0.3172	0.2858
0.8523	10.7914	6000	0.3189	0.2857
0.8602	11.5108	6400	0.3075	0.2850
0.811	12.2302	6800	0.2993	0.2833
0.7986	12.9496	7200	0.2959	0.2836
0.7744	13.6691	7600	0.2946	0.2816
0.7686	14.3885	8000	0.2959	0.2837
0.7379	15.1079	8400	0.2926	0.2814
0.7554	15.8273	8800	0.2890	0.2817
0.7386	16.5468	9200	0.2876	0.2799
0.7128	17.2662	9600	0.2861	0.2821
0.7063	17.9856	10000	0.2870	0.2826
0.704	18.7050	10400	0.2822	0.2806
0.6876	19.4245	10800	0.2904	0.2831
0.6898	20.1439	11200	0.2781	0.2830
0.694	20.8633	11600	0.2814	0.2818
0.683	21.5827	12000	0.2783	0.2803
0.6871	22.3022	12400	0.2774	0.2805
0.6708	23.0216	12800	0.2794	0.2795
0.6641	23.7410	13200	0.2759	0.2808
0.6655	24.4604	13600	0.2729	0.2808
0.6626	25.1799	14000	0.2753	0.2812
0.6481	25.8993	14400	0.2756	0.2802
0.6404	26.6187	14800	0.2749	0.2816
0.6579	27.3381	15200	0.2733	0.2805
0.6352	28.0576	15600	0.2731	0.2804
0.6517	28.7770	16000	0.2729	0.2810
0.6495	29.4964	16400	0.2725	0.2806

Framework versions

• Transformers 4.57.6
• Pytorch 2.9.1+cu128
• Datasets 4.5.0
• Tokenizers 0.22.2