wav2vec2-lv-60-espeak-cv-ft-custom_vocab-OtherDiacritics-ds-f7

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: 0.2194
• Per: 0.2658

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
17.6605	0.7194	400	4.5622	0.9860
4.3982	1.4388	800	4.0924	0.9860
4.0852	2.1583	1200	3.6801	0.9860
3.289	2.8777	1600	1.5881	0.4300
2.0883	3.5971	2000	0.8783	0.3427
1.6008	4.3165	2400	0.6054	0.3077
1.3534	5.0360	2800	0.4754	0.2917
1.2201	5.7554	3200	0.4041	0.2852
1.1139	6.4748	3600	0.3566	0.2813
1.0303	7.1942	4000	0.3300	0.2768
0.9654	7.9137	4400	0.3140	0.2749
0.9383	8.6331	4800	0.2941	0.2740
0.8962	9.3525	5200	0.2812	0.2733
0.8665	10.0719	5600	0.2777	0.2725
0.8579	10.7914	6000	0.2695	0.2707
0.8512	11.5108	6400	0.2628	0.2712
0.8363	12.2302	6800	0.2579	0.2714
0.791	12.9496	7200	0.2485	0.2699
0.7716	13.6691	7600	0.2472	0.2689
0.7773	14.3885	8000	0.2409	0.2692
0.7443	15.1079	8400	0.2438	0.2693
0.736	15.8273	8800	0.2395	0.2685
0.7459	16.5468	9200	0.2381	0.2686
0.7273	17.2662	9600	0.2420	0.2700
0.711	17.9856	10000	0.2347	0.2689
0.6997	18.7050	10400	0.2348	0.2682
0.6989	19.4245	10800	0.2315	0.2682
0.7056	20.1439	11200	0.2310	0.2669
0.6756	20.8633	11600	0.2304	0.2676
0.6873	21.5827	12000	0.2304	0.2682
0.6999	22.3022	12400	0.2250	0.2671
0.6735	23.0216	12800	0.2232	0.2672
0.6835	23.7410	13200	0.2229	0.2671
0.6648	24.4604	13600	0.2202	0.2655
0.6612	25.1799	14000	0.2199	0.2658
0.6477	25.8993	14400	0.2215	0.2669
0.6389	26.6187	14800	0.2217	0.2658
0.6742	27.3381	15200	0.2206	0.2663
0.6299	28.0576	15600	0.2208	0.2664
0.6369	28.7770	16000	0.2207	0.2661
0.6458	29.4964	16400	0.2194	0.2658

Framework versions

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