fl_asr_speech_recognition

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2947
• Wer: 0.1449
• Cer: 0.0451

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: 0.0001
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 200
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.625	12.6582	1000	0.4832	0.5625	0.1090
0.3037	25.3165	2000	0.3879	0.3665	0.0686
0.2127	37.9747	3000	0.4096	0.2926	0.0617
0.1767	50.6329	4000	0.3967	0.25	0.0552
0.1238	63.2911	5000	0.3024	0.2273	0.0529
0.0868	75.9494	6000	0.3768	0.2330	0.0487
0.0823	88.6076	7000	0.2742	0.2244	0.0420
0.0696	101.2658	8000	0.2792	0.2074	0.0383
0.0496	113.9241	9000	0.3362	0.1591	0.0359
0.0413	126.5823	10000	0.3061	0.1562	0.0400
0.0286	139.2405	11000	0.3264	0.1591	0.0406
0.0294	151.8987	12000	0.3046	0.1648	0.0424
0.0183	164.5570	13000	0.3083	0.1506	0.0400
0.0159	177.2152	14000	0.2947	0.1449	0.0451
0.009	189.8734	15000	0.3198	0.1477	0.0411

Framework versions

• Transformers 4.43.3
• Pytorch 2.3.1+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1