240803_01-wav2vec2-ASR-Global-All-Clients

This model is a fine-tuned version of zainulhakim/240626-wav2vec2-ASR_Global on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2829
• Wer: 0.1449

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: 5
• 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: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	1.5625	100	0.2964	0.1989
No log	3.125	200	0.3052	0.2216
No log	4.6875	300	0.3365	0.2614
No log	6.25	400	0.3643	0.2670
0.1191	7.8125	500	0.3596	0.2841
0.1191	9.375	600	0.4020	0.2358
0.1191	10.9375	700	0.3445	0.2727
0.1191	12.5	800	0.3989	0.3239
0.1191	14.0625	900	0.3855	0.2415
0.1639	15.625	1000	0.3188	0.2131
0.1639	17.1875	1100	0.2902	0.1648
0.1639	18.75	1200	0.2829	0.1449

Framework versions

• Transformers 4.42.4
• Pytorch 2.3.1+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1