wav2vec2-E10_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: 2.3467
• Cer: 50.0764

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
33.121	0.1289	200	5.0379	100.0
5.0221	0.2579	400	4.6710	100.0
4.8585	0.3868	600	4.6327	100.0
4.8155	0.5158	800	4.6354	100.0
4.7249	0.6447	1000	4.6274	98.6313
4.6786	0.7737	1200	4.5950	99.5418
4.6106	0.9026	1400	4.5545	97.9382
4.4195	1.0316	1600	4.4562	99.2364
4.1478	1.1605	1800	4.2379	97.5623
3.8234	1.2895	2000	3.9072	78.5714
3.39	1.4184	2200	3.5827	76.3628
3.2255	1.5474	2400	3.4255	73.1497
2.9972	1.6763	2600	3.1945	66.3593
2.8726	1.8053	2800	3.0039	62.3884
2.7321	1.9342	3000	2.9409	62.7526
2.6263	2.0632	3200	2.7179	58.0945
2.5459	2.1921	3400	2.6416	57.8536
2.4325	2.3211	3600	2.6501	56.4850
2.3828	2.4500	3800	2.5066	55.4688
2.2717	2.5790	4000	2.4522	53.1837
2.2597	2.7079	4200	2.3933	50.6403
2.2042	2.8369	4400	2.3427	50.8752
2.197	2.9658	4600	2.3467	50.0764

Framework versions

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