wav2vec2-E30_freq_speed_pause

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.0912
• Cer: 46.1231

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: 8
• 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: 50
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
31.7937	0.1289	200	5.1147	100.0
4.9653	0.2579	400	4.6684	100.0
4.8114	0.3868	600	4.6765	100.0
4.7658	0.5158	800	4.6123	97.7150
4.6791	0.6447	1000	4.6076	98.9544
4.6438	0.7737	1200	4.6205	97.6974
4.5903	0.9026	1400	4.4614	97.8442
4.439	1.0316	1600	4.4028	98.2848
4.1968	1.1605	1800	4.2323	94.1612
3.8917	1.2895	2000	3.8326	78.5127
3.5148	1.4184	2200	3.6092	70.7119
3.2601	1.5474	2400	3.3938	71.4873
3.0276	1.6763	2600	3.1059	64.2094
2.8883	1.8053	2800	2.9391	61.2841
2.7381	1.9342	3000	2.7814	59.1929
2.5905	2.0632	3200	2.5964	54.9988
2.4555	2.1921	3400	2.3926	51.0456
2.3566	2.3211	3600	2.3930	51.1689
2.2751	2.4500	3800	2.2846	49.4596
2.1796	2.5790	4000	2.1934	48.0028
2.1292	2.7079	4200	2.1426	47.0923
2.0724	2.8369	4400	2.1201	47.0042
2.0759	2.9658	4600	2.0912	46.1231

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.19.1