wav2vec2-lv-60-espeak-cv-ft-custom_vocab-OtherDiacritics-ds-f2

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: 0.2691
• Per: 0.2744

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
17.7607	0.7194	400	4.5613	0.9850
4.4044	1.4388	800	4.0840	0.9850
4.0699	2.1583	1200	3.6606	0.9850
3.2389	2.8777	1600	1.5561	0.4345
2.0542	3.5971	2000	0.8710	0.3440
1.5827	4.3165	2400	0.6341	0.3114
1.352	5.0360	2800	0.4889	0.2939
1.2081	5.7554	3200	0.4297	0.2884
1.1192	6.4748	3600	0.3864	0.2869
1.0281	7.1942	4000	0.3653	0.2859
0.9994	7.9137	4400	0.3475	0.2835
0.9303	8.6331	4800	0.3358	0.2819
0.8754	9.3525	5200	0.3202	0.2824
0.8718	10.0719	5600	0.3114	0.2802
0.8346	10.7914	6000	0.3123	0.2787
0.8321	11.5108	6400	0.3063	0.2799
0.818	12.2302	6800	0.3025	0.2799
0.7748	12.9496	7200	0.2995	0.2772
0.7789	13.6691	7600	0.2962	0.2770
0.7745	14.3885	8000	0.2965	0.2786
0.7377	15.1079	8400	0.2898	0.2758
0.7354	15.8273	8800	0.2926	0.2764
0.7333	16.5468	9200	0.2885	0.2768
0.7084	17.2662	9600	0.2859	0.2772
0.7146	17.9856	10000	0.2832	0.2744
0.6901	18.7050	10400	0.2833	0.2755
0.695	19.4245	10800	0.2860	0.2758
0.6795	20.1439	11200	0.2813	0.2751
0.675	20.8633	11600	0.2769	0.2739
0.6778	21.5827	12000	0.2766	0.2746
0.6734	22.3022	12400	0.2761	0.2757
0.6602	23.0216	12800	0.2713	0.2751
0.6633	23.7410	13200	0.2721	0.2747
0.6686	24.4604	13600	0.2683	0.2744
0.6388	25.1799	14000	0.2683	0.2740
0.6695	25.8993	14400	0.2674	0.2740
0.6453	26.6187	14800	0.2682	0.2737
0.6552	27.3381	15200	0.2699	0.2743
0.6273	28.0576	15600	0.2710	0.2744
0.6399	28.7770	16000	0.2690	0.2739
0.6349	29.4964	16400	0.2691	0.2744

Framework versions

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