wav2vec2-lv-60-espeak-cv-ft-custom_vocab-OtherDiacritics-ds-f5

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: 0.2546
• Per: 0.2734

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
17.6175	0.7194	400	4.5662	0.9888
4.4006	1.4388	800	4.0899	0.9888
4.0709	2.1583	1200	3.6440	0.9886
3.2766	2.8777	1600	1.5594	0.4415
2.0891	3.5971	2000	0.8822	0.3426
1.5968	4.3165	2400	0.6211	0.3105
1.3795	5.0360	2800	0.4977	0.2938
1.1989	5.7554	3200	0.4303	0.2885
1.1349	6.4748	3600	0.3811	0.2837
1.0385	7.1942	4000	0.3563	0.2818
0.994	7.9137	4400	0.3444	0.2828
0.9319	8.6331	4800	0.3272	0.2808
0.8814	9.3525	5200	0.3135	0.2789
0.8881	10.0719	5600	0.3031	0.2769
0.8555	10.7914	6000	0.3001	0.2775
0.8536	11.5108	6400	0.2988	0.2789
0.8036	12.2302	6800	0.2932	0.2775
0.7946	12.9496	7200	0.2906	0.2776
0.7722	13.6691	7600	0.2869	0.2785
0.7714	14.3885	8000	0.2855	0.2775
0.7507	15.1079	8400	0.2792	0.2761
0.7522	15.8273	8800	0.2845	0.2771
0.7436	16.5468	9200	0.2746	0.2754
0.7154	17.2662	9600	0.2755	0.2771
0.7119	17.9856	10000	0.2711	0.2753
0.6945	18.7050	10400	0.2720	0.2749
0.6855	19.4245	10800	0.2671	0.2744
0.6885	20.1439	11200	0.2634	0.2743
0.683	20.8633	11600	0.2645	0.2742
0.6752	21.5827	12000	0.2660	0.2744
0.678	22.3022	12400	0.2592	0.2726
0.6502	23.0216	12800	0.2581	0.2730
0.6784	23.7410	13200	0.2585	0.2740
0.665	24.4604	13600	0.2580	0.2736
0.6681	25.1799	14000	0.2558	0.2728
0.6448	25.8993	14400	0.2534	0.2735
0.6231	26.6187	14800	0.2523	0.2728
0.6542	27.3381	15200	0.2540	0.2731
0.6349	28.0576	15600	0.2558	0.2735
0.6492	28.7770	16000	0.2547	0.2735
0.6252	29.4964	16400	0.2546	0.2734

Framework versions

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