wav2vec2-E50_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: 4.2983
• Cer: 85.6874

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
37.7103	0.1289	200	6.7374	100.0
15.4444	0.2579	400	4.8863	94.0188
10.7358	0.3868	600	4.6351	93.4371
10.2187	0.5158	800	4.6612	92.4736
9.4989	0.6447	1000	4.5912	92.1680
8.4222	0.7737	1200	4.9206	91.7274
6.6021	0.9026	1400	4.4301	91.6275
4.7756	1.0316	1600	4.4769	93.6428
4.5225	1.1605	1800	4.4304	93.2902
4.4805	1.2895	2000	4.4101	92.9318
4.4489	1.4184	2200	4.3892	92.9788
4.4476	1.5474	2400	4.4033	92.7438
4.4293	1.6763	2600	4.3643	92.8202
4.4077	1.8053	2800	4.3827	92.7438
4.3843	1.9342	3000	4.3231	92.9083
4.3661	2.0632	3200	4.3462	89.3361
4.3493	2.1921	3400	4.4030	88.2197
4.3051	2.3211	3600	4.3176	87.6146
4.2924	2.4500	3800	4.2757	85.8696
4.2824	2.5790	4000	4.3263	83.2726
4.2591	2.7079	4200	4.3128	84.9060
4.2231	2.8369	4400	4.3405	85.1821
4.1965	2.9658	4600	4.2983	85.6874

Framework versions

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