wav2vec2-Y_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: 3.5699
• Cer: 73.6313

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
32.2129	0.1289	200	5.0008	100.0
5.033	0.2579	400	4.7905	100.0
4.8386	0.3868	600	4.7426	100.0
4.8071	0.5158	800	4.6807	100.0
4.7063	0.6447	1000	4.6552	97.3625
4.66	0.7737	1200	4.5850	96.8985
4.5669	0.9026	1400	4.5298	97.6739
4.355	1.0316	1600	4.5565	96.8574
4.1082	1.1605	1800	4.4688	95.0012
3.8409	1.2895	2000	4.5259	80.5804
3.4814	1.4184	2200	4.2544	80.0164
3.3101	1.5474	2400	4.3252	79.2822
3.1423	1.6763	2600	4.1301	76.4979
3.0633	1.8053	2800	4.1739	75.8694
2.9237	1.9342	3000	4.1099	75.8870
2.8312	2.0632	3200	4.0269	77.0148
2.7206	2.1921	3400	3.8298	76.4333
2.6076	2.3211	3600	3.7397	75.5698
2.5695	2.4500	3800	3.7239	75.1586
2.4758	2.5790	4000	3.6991	74.1071
2.4508	2.7079	4200	3.5936	73.6725
2.3894	2.8369	4400	3.5976	73.7136
2.3974	2.9658	4600	3.5699	73.6313

Framework versions

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