ASR-cv-corpus-ug-13

This model is a fine-tuned version of piyazon/ASR-cv-corpus-ug-12 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0048
• Wer: 0.0049

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: 4e-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: 400
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0054	0.2600	500	0.0119	0.0163
0.0102	0.5200	1000	0.0102	0.0129
0.0096	0.7800	1500	0.0101	0.0162
0.0081	1.0400	2000	0.0068	0.0100
0.0062	1.3001	2500	0.0096	0.0128
0.0071	1.5601	3000	0.0076	0.0095
0.0067	1.8201	3500	0.0086	0.0122
0.0066	2.0801	4000	0.0072	0.0103
0.0048	2.3401	4500	0.0080	0.0115
0.005	2.6001	5000	0.0080	0.0110
0.0055	2.8601	5500	0.0081	0.0109
0.0047	3.1201	6000	0.0063	0.0092
0.0034	3.3801	6500	0.0066	0.0087
0.0042	3.6401	7000	0.0079	0.0105
0.0041	3.9002	7500	0.0061	0.0085
0.0035	4.1602	8000	0.0080	0.0109
0.003	4.4202	8500	0.0088	0.0118
0.0036	4.6802	9000	0.0069	0.0093
0.0023	4.9402	9500	0.0060	0.0087
0.0024	5.2002	10000	0.0064	0.0082
0.0019	5.4602	10500	0.0076	0.0086
0.0021	5.7202	11000	0.0062	0.0086
0.0022	5.9802	11500	0.0055	0.0078
0.0013	6.2402	12000	0.0065	0.0081
0.0016	6.5003	12500	0.0063	0.0076
0.0013	6.7603	13000	0.0050	0.0072
0.0012	7.0203	13500	0.0056	0.0067
0.001	7.2803	14000	0.0052	0.0063
0.0008	7.5403	14500	0.0049	0.0061
0.0004	7.8003	15000	0.0049	0.0053
0.0007	8.0603	15500	0.0051	0.0055
0.0003	8.3203	16000	0.0052	0.0055
0.0004	8.5803	16500	0.0051	0.0053
0.0004	8.8404	17000	0.0048	0.0051
0.0003	9.1004	17500	0.0049	0.0050
0.0002	9.3604	18000	0.0048	0.0050
0.0002	9.6204	18500	0.0048	0.0049
0.0002	9.8804	19000	0.0048	0.0049

Framework versions

• Transformers 4.56.1
• Pytorch 2.8.0+cu128
• Datasets 4.1.0
• Tokenizers 0.22.0