wav2vec2-E10_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: 2.1732
• Cer: 46.9807

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
34.0782	0.1289	200	5.1516	100.0
5.0781	0.2579	400	4.7666	100.0
4.9696	0.3868	600	4.6840	100.0
4.906	0.5158	800	4.6904	100.0
4.7963	0.6447	1000	4.6462	100.0
4.7565	0.7737	1200	4.6231	100.0
4.7372	0.9026	1400	4.5947	100.0
4.6765	1.0316	1600	4.5636	100.0
4.6162	1.1605	1800	4.5811	99.9354
4.4587	1.2895	2000	4.4613	96.3346
4.1506	1.4184	2200	4.2476	95.7766
3.8599	1.5474	2400	3.8368	73.1262
3.34	1.6763	2600	3.3885	64.8085
3.1038	1.8053	2800	3.1215	61.0491
2.9058	1.9342	3000	2.9318	58.3705
2.7231	2.0632	3200	2.7934	57.7244
2.5775	2.1921	3400	2.6023	53.1661
2.4413	2.3211	3600	2.5756	54.5759
2.3662	2.4500	3800	2.3468	50.9810
2.2503	2.5790	4000	2.2820	49.2892
2.231	2.7079	4200	2.2980	49.1483
2.1522	2.8369	4400	2.1792	47.4389
2.1532	2.9658	4600	2.1732	46.9807

Framework versions

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