wav2vec2-E10_freq_speed_pause2

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: 1.2939
• Cer: 29.2538

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: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• 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
33.6216	0.1289	200	5.0765	100.0
5.0003	0.2579	400	4.7569	100.0
4.8152	0.3868	600	4.6179	97.9730
4.7689	0.5158	800	4.6068	97.8966
4.6602	0.6447	1000	4.5302	99.8237
4.4084	0.7737	1200	4.4900	97.2150
3.6156	0.9026	1400	3.2667	63.0787
2.9009	1.0316	1600	2.8830	57.7027
2.621	1.1605	1800	2.9525	52.9083
2.3029	1.2895	2000	2.4458	49.5593
2.046	1.4184	2200	2.0830	41.4982
1.8452	1.5474	2400	1.9670	40.0940
1.6929	1.6763	2600	1.8540	38.7955
1.5456	1.8053	2800	1.6959	36.3690
1.4492	1.9342	3000	1.6960	36.4219
1.3279	2.0632	3200	1.5604	34.0541
1.1872	2.1921	3400	1.4687	33.3255
1.1379	2.3211	3600	1.4123	31.0517
1.071	2.4500	3800	1.3426	29.1657
1.0709	2.5790	4000	1.3340	29.9412
1.0125	2.7079	4200	1.3153	29.4007
0.9648	2.8369	4400	1.3055	29.9706
0.9444	2.9658	4600	1.2939	29.2538

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3